Underpricing, share retention, and the IPO aftermarket liquidity Mingsheng Li* University of Louisiana at Monroe, Monroe, LA 71209, U.S.A. Steven Xiaofan Zheng University of Manitoba, Winnipeg, MB, Canada R3T 2N2 Melissa V. Melancon University of Louisiana at Monroe, Monroe, LA 71209, U.S.A.
Current version: March 2005 Key words: Initial public offerings; share retention; liquidity; spreads. JEL classification: G12, G14
*Corresponding author: Mingsheng Li, Department of Economics and Finance, College of Business Administration, University of Louisiana at Monroe, Monroe, LA 71209. Phone: 3183421169; Fax: 3183423096; Email:
[email protected]
Underpricing, share retention, and the IPO aftermarket liquidity
Research paper Purpose To test the effects of underpricing and share retention (i.e., the proportion of shares retained by the preIPO owners) on IPO aftermarket liquidity. Design/methodology/approach We use both percentage spread and turnover ratio to measure liquidity. The percentage spread is the quoted bidask spread divided by the quoted midpoint and measures the trading cost relative to share price. Turnover ratio is the daily trading volume divided by the number of shares offered and measures the speed of transaction. We conduct both nonparametric analyses and multiple regressions to investigate the effects of underpricing and share retention on liquidity. Findings Our results indicate that initial return is positively related to turnover ratio and negatively related to percentage spread. These relations are significant even after controlling for other factors. We also find that the preIPO owners’ retention rate is positively related to turnover ratio and negatively related to percentage spread. High retention rates attract more trades, provide quality assurance, and improve IPO aftermarket liquidity. Originality/value This paper investigates the theoretical links between underpricing and liquidity and provides direct evidence on Booth and Chua’s (1996) liquidity theory. In addition, this is one of the first empirical studies to analyze the effect of share retention on aftermarket liquidity.
1
1. Introduction The underpricing of initial public offerings (IPOs) and the proportion of shares retained by the preIPO owners (share retention) are two important decision variables in the IPO process that have attracted much attention in the IPO literature.1 However, the effects of these factors on aftermarket liquidity are not clear. Although liquidity theory suggests that underpricing increases the IPO aftermarket liquidity, there is little direct evidence on how the two are related. Further, studies on how share retention affects market liquidity are rather limited. In this study, we investigate the theoretical links between underpricing, share retention, and liquidity to provide direct evidence on this issue. According to Booth and Chua (1996), issuers’ demand for a liquid aftermarket creates incentives for underpricing. They argue that oversubscription for a new issue induces broad initial ownership, which in turn increases secondary market liquidity. However, broad initial ownership increases investorborne information costs, which must be offset through underpricing. Therefore, they suggest that underpricing is a positive function of ownership dispersion and secondarymarket liquidity. In other words, the positive relation between underpricing and aftermarket liquidity depends on three links: (1) a positive relation between underpricing and oversubscription for the new issues; (2) a positive relation between oversubscription and ownership dispersion; and (3) a positive relation between ownership dispersion and market liquidity. Intuitively, underpricing and oversubscription are positively related, because the more underpriced a new issue is, the more investors it attracts. This is consistent with Benveniste and Spindt’s (1989) view that underwriters underprice new issues to induce investors to reveal private information and to attract investors’ interests in the IPOs. Koh and Walter (1989) and Brennan and Franks (1997) provide empirical evidence supporting the positive relation between underpricing and oversubscription. The positive relation between ownership dispersion and 1
See Ritter and Welch (2002) for a detailed literature review.
2
liquidity is well documented in the literature (Merton, 1987; Bhushan, 1989). Thus, the validity of the positive relation between underpricing and aftermarket liquidity largely depends on whether oversubscription could lead to greater ownership dispersion. Neither theoretical nor empirical studies fully support the positive relation between oversubscription and ownership dispersion. Once the total number of shares offered is determined, the initial ownership structure largely depends on how shares are allocated. In the United States, underwriters have considerable latitude in share allocation. To reduce the likelihood of monitoring by blockholders, issuers and underwriters could use oversubscription to discriminate against these investors, which would lead to a more dispersed ownership (Brennan and Franks, 1997). Conversely, underwriters could also use oversubscription to favor blockholders, since the monitoring service provided by blockholders reduces agency costs and increases firm value (Stoughton and Zechner, 1998). Field and Sheehan (2004) find that the link between underpricing and ownership structure is weak; hence, how underpricing relates to aftermarket liquidity is an empirical question. This study provides direct evidence on this issue. Pham, Kalev, and Steen (2003) investigate underpricing, stock allocation, ownership structure, and aftermarket liquidity using 113 Australian IPOs. They find evidence supporting the positive relation between underpricing, shareholder distribution, and aftermarket liquidity. However, this finding is not applicable to IPOs in the United States for two critical reasons. First, Australian procedures for issuing new shares differ markedly from the procedures employed in the United States (Lee, Taylor, and Walter, 1996). Australian firms predominantly use fixedpricing, which does not allow for changes in either the offer price or the number of shares offered after the prospectus is filed. In contrast, U.S. firms use bookbuilding, in which the offer price and the number of shares offered can be altered until the final offer price is determined. Among the 113 IPOs studied by Pham, Kalev, and Steen, only four firms adopt the bookbuilding process.
3
Second, Australian institutional arrangements differ substantially from those in the United States. The U.S. underwriters’ role is much broader than underwriting and extends to the aftermarket; U.S. underwriters undertake different actions to stabilize the aftermarket price, but these activities are illegal in Australia (Bayley, Lee, and Walter, 2003). While investors in Australia may freely sell their allocated shares, U.S. underwriters discourage and punish investors who sell their allocated shares immediately (flipping) in the aftermarket. These differences directly affect the aftermarket liquidity. How share retention affects liquidity is far less clear. According to Zheng, Ogden, and Jen (2003), underpricing boosts liquidity, especially when the proportion of shares retained by preIPO owners is large. They argue that when more shares are retained (i.e., the number of shares floating in the market is reduced), a stock’s liquidity declines. However, this sharefloating effect is only one of the possible aspects related to share retention. Examining other factors helps define the relation between share retention and aftermarket liquidity. A higher retention rate by preIPO owners limits the number of shares available for trade, but it does not necessarily lead to lower liquidity. The signaling theory of Leland and Pyle (1977) suggests that the proportion of equity retained by the original owners is positively related to firm value. Because a less diversified portfolio is more costly to maintain, entrepreneurs retain significant ownership interests only if future cash flow is expected to be high. Consistent with signaling theory, Downes and Heinkel (1982) and Jain and Kini (1994) find a significant positive relation between postIPO operation performance and the proportion of equity retained by the preIPO owners. According to the signaling theory and empirical evidence enumerated above, the current study posits that high retention by preIPO owners will attract more investors in the secondary market and increase market liquidity, ceteris paribus. However, the ultimate relation between share retention and secondarymarket liquidity depends on whether the sharefloating effect or the signaling effect dominates. If the sharefloating effect is dominant, high retention reduces market
4
liquidity. In contrast, if the signaling effect is the key to secondarymarket liquidity, high retention increases market liquidity. This study provides direct empirical evidence on this interesting issue. We select 1,673 IPOs of common stocks listed on NASDAQ between 1993 and 2000. We summarize the following main findings. First, initial return (underpricing) is positively related to turnover ratio and negatively related to percentage spreads. The relation is significant after controlling for other factors. This indicates that initial return is a positive function of liquidity. Second, the proportion of shares retained by the preIPO owners is positively related to turnover ratio and negatively related to percentage spread. Consistent with signaling theory, this suggests that a high retention rate sends a positive market signal, which attracts more trades and improves market liquidity. Third, consistent with previous studies, we find that market liquidity is also affected by other factors such as share price, firm size, return volatility, and reduction of tick size.
2. Data, sample selection, and offering characteristics We collect firm names (ticker symbols), offer dates, offer prices, the number of shares offered, the identities of book runners, and backing of venture capitalists from the Securities Data Company’s (SDC) database.
Daily insider trading volume is obtained from the Thomson
Financial Net work (TFN). We also collect share codes, share types, the number of shares outstanding, industry codes (SIC), and listing exchanges from the Center for Research in Security Prices (CRSP). Intraday minutebyminute trades and quotes are retrieved from the Trade and Quotes database (TAQ). The SDC database lists more than 2,690 IPOs that came to the market from 1993 to 2000. From these new issues, we select 2,256 IPOs listed on NASDAQ with offer prices greater than or
5
equal to $5.2 We exclude: (1) companies incorporated outside the United States, closedend funds, unit offerings, and Real Estate Investment trusts; (2) firms that have less than 12 months of data from either CRSP or TAQ after the IPO offering date; and (3) firms that changed listings or experienced mergers or acquisitions within oneyear after the offerings. After applying these filters, there are 1,673 IPOs left in the sample. For the intraday quote data, we exclude observations with: (1) quotes before opening (9:30) or after closing (16:00); (2) bid or ask prices ≤ 0; (3) quoted spread at bid and ask prices >$5 or ≤ 0; (4) the ratio of bid price or ask price to the quoted midpoint >0.5; and (5) nonfirm quotes and quotes corresponding to trading halts. For trade data, we eliminate observations with: (1) trades before opening (9:30) or after closing (16:00); (2) trades reported out of time sequence; and (3) cancelled trades. These observations are deleted because these quotes do not represent the normal quotes or are due to errors. Table 1 summarizes the offering statistics of the IPO sample. The mean and median prices are $11.94 and $12 for the whole sample, indicating a uniform distribution around the mean. The initial return, which is calculated as [the first day closing price  offer price]/offer price, is about 23.54% on average for the whole sample, with a median of 11.76%. The initial return in recent years (1999 and 2000) ranges from 44% to 53%, which is much higher than that in earlier periods, supporting the Dot.com bubble documented in Ljungqvist and Wilhelm (2003).3 The number of shares offered is also significantly larger in recent years. The proportion 2
There are two main reasons for us to exclude IPOs listed on NYSE/Amex. First, the number of IPOs listed
on NYSE/Amex is much smaller than that on NASDAQ. Second, spreads are quoted differently for stocks traded on NYSE/Amex and NASDAQ due to different trading mechanisms. To make spread measures comparable, we use only the Nasdaq IPO sample. 3
The average initial return reported in this study for the IPOs in 1999 and 2000 is smaller than that
reported in Ritter and Welch (2002). Two factors may explain this disparity. First, we exclude IPOs that experienced mergers or acquisitions or those that changed exchanges before the first anniversary of
6
of shares retained by the preIPO owners is 68% on average, which is relatively stable during the test period.
3. Underpricing and aftermarket liquidity Liquidity has multiple dimensions and could be better measured by the price concessions needed for an immediate transaction and the speed of transaction (Demsetz, 1968; Stoll, 2000). Following the microstructure literature, we adopt the two most commonly used liquidity measures: percentage spread and turnover ratio. The percentage spreadthe quoted spread divided by the quoted midpointmeasures the trading cost relative to share price. The larger the spreads are, the lower the liquidity is. Although many studies also use quoted spreads and effective spreads to measure liquidity, these measures do not control for prices. In relative terms, a $0.10 spread for a $10 stock is much larger than a $0.20 spread for a $40 stock. Hence, we believe that percentage spread is a more relevant measure than quoted or effective spreads. The second measure of liquidity we use is turnover ratio, which is the daily trading volume divided by shares offered and measures the speed of transaction. A large turnover ratio indicates high liquidity.
3.1 Liquidity in the earlier period of the IPO aftermarket To test Booth and Chua’s (1996) hypothesis that underpricing is positively related to liquidity, we divide the sample IPOs into four groups based on the initial return. Group 1 (4)
offering. Prior studies (Schultz and Zaman, 2001) find that many Internet firms experience mergers and acquisitions shortly after the offering. Thus, the IPOs excluded from our sample are likely to be those with large initial returns. Second, our sample includes only IPOs listed on NASDAQ. This may also affect our results.
7
includes stocks with the smallest (largest) initial return.4 For each stock, we first calculate the mean values of the two liquidity measures in a 20trading day period after the offerings. Then we compute and compare the group averages across the four initial return groups. To analyze whether liquidity has an increasing or a decreasing trend across the initial return groups, we conduct nonparametric JonckheereTerpstra (JT) tests since the liquidity measures are truncated at zero. For increasing trend analysis, the null hypothesis is Ho: X1=X2=X3=X4, against the alternative hypothesis H1: X1X4. Table 2, Panel A1 summarizes the results in the first 20tradingday period. The average initial return is 23.55% for the whole sample, ranging from –2.24% for group 1 to 71.06% for group 4. The difference among groups is significant at the 0.01 level. Percentage spread is inversely related to initial return. The percentage spread is 2.88% for the whole sample and declines from 3.22% for group 1 to 2.06% for group 4. The JonckheereTerpstra test supports the alternative hypothesis that initial return is inversely related to the percentage spread, with a JTstatistic of –13.09. In contrast, turnover ratio is positively related to initial return. The average turnover ratio for the whole group during the first 20tradingday period is 9.38%. It increases from 6.85% for group 1 to 15.10% for group 4. The JonckheereTerpstra test shows that the increasing trend is significant, with a JTstatistic of 21.2. 5 4
Ljungqvist and Wilhelm (2003) find that IPOs in the bubble period (1999 and 2000) have larger initial
returns than IPOs in other years. Thus, more IPOs in the bubble period are in the large initial return groups. To reduce this time effect, we also divide IPOs into four classes based on the initial return in the year they came to the market and replicate all the analyses. The results are essentially unchanged. 5
Some studies exclude the first few days when calculating turnover ratio due to the large trading volume
during this period. When we conduct the analysis by excluding the first three trading days, the differences among the four groups are essentially the same, although the average turnover ratio is lower than that
8
These preliminary results support Booth and Chua’s (1996) liquidity hypothesis, and are consistent with the results of Australian IPOs studied by Pham, Kalev, and Steen (2003). However, these results may simply reflect the effect of stock price on liquidity rather than the effect of initial return, since empirical evidence indicates that highprice stocks have smaller percentage spreads (McInish and Wood, 1992; Christie and Schultz, 1994; Stoll, 2000). As a preliminary test of the effect of price on spreads, we compare the aftermarket price among the initial return groups. To avoid trade price bounce between bid and ask prices, we use the average of quoted midpoint. The average quoted midpoint for the whole sample is $15.72 during the first 20tradingday period of the aftermarket. It increases from $10.49 for group 1 to $25.08 for group 4. The positive relation is both statistically and economically significant at the 0.01 level (JTstatistic = 26.97). Thus, it is quite possible that the inverse relation between initial return and percentage spreads is driven by stock price. We will provide additional evidence in later sections.
3.2 Liquidity in a 240tradingday period Studies show that many provisions of Securities and Exchange Commission (SEC) regulations and underwriter restrictions are implemented in the earlier period of the IPO aftermarket. These provisions and restrictions affect spreads and market liquidity. For example, Aggarwal (2000) and Ellis, Michaely, and O’Hara (2000) find that most price supports are carried out during the first 20 days after offerings. The quiet period mandated by the SEC lasts for 25 calendar days, and penalty bids implemented by the underwriters are lifted after 30 days. Thus, spreads in the first 20 trading days may not reflect the overall aftermarket liquidity. For these reasons, we compare percentage spread and turnover ratio over a 240tradingday period (approximately one calendar year) and report the results in Panel A2 of Table 2. The
including the first three days. We prefer including all days in our analysis since the first three days are an important component of the aftermarket.
9
percentage spread is 3.3% for the whole sample and declines from 3.77% for group 1 to 2.51% for group 4. The decreasing trend is significant, with a JTstatistic of 13.05. Also, note that spreads are larger during the 240tradingday period than during the first 20tradingday period. Nevertheless, the decreasing trend is similar to that in the first 20tradingday period. The larger spreads in the 240tradingday period are consistent with Hegde and Miller’s (1989) finding that bidask spreads for IPOs are about threefourths as large as the bidask spreads of seasoned equities in the early aftermarket. The turnover ratio is 3.7% for the whole sample in the 240tradingday period. It increases from 2.81% for group 1 to 5.82% for group 4. The quoted midpoint for group 4 is also significantly higher than that of the other three groups. As an additional test, we also exclude the IPOs during the bubble period (1999 and 2000), since the IPOs during this unusual period could skew the results. The results are reported in Panel B of Table 2. The initial return for this subsample is 17.16%, which is smaller than the whole sample. The liquidity measures in both the first 20 tradingday period (Panel B1) and the 240tradingday period (Panel B2) are similar to those of the whole sample. This suggests that the observed relation between initial return and liquidity is not driven by the IPOs in the bubble period. For this reason, we use the whole sample for all other analyses. Although not reported, we repeat this process over 12 20tradingday periods. The variations in spreads across the initial return groups are qualitatively unchanged. In addition, we compare percentage spread across the four initial return groups in a 20tradingday window before and after the lockup. The cross sectional variations in spreads are essentially the same. For expositional purposes, we focus on the results in the 20 and 240tradingday periods.6
6
Two important events that significantly affected spreads occurred on the NASDAQ during our test period.
The Order Handling Rule (OHR) was implemented on a phasedin basis, beginning on January 20, 1997 and ending on October 13, 1997. The tick size reduction became effective on June 2, 1997. The minimum tick size changed from $1/8 to $1/16 for stocks selling at prices equal to or greater than $10. To capture the
10
3.3 Comparing spreads after controlling for aftermarket share price To test whether the observed relation between initial return and spreads is driven merely by share price, we use univariate analysis comparing spreads by controlling for share price. We first divide the whole sample into four groups based on share price in a 240tradingday period after the IPO. Then, we divide stocks in each of the price group into four subgroups based on the initial return. If price is a dominant factor in the observed variation in spreads, significant differences will not exist across the initial return subgroups after controlling for price. In contrast, if initial return plays a key role in the spread pattern, differences in spreads will remain after controlling for price. Table 3 reports the results. After controlling for price, initial return is still negatively correlated with percentage spread. However, the inverse relation is significant only for the higherprice groups (groups 3 and 4). The positive relation between turnover ratio and initial return remains significant for each of the four price groups. Taking price group 1 (Panel A1) as an example, the turnover ratio is 2.56% for the smallest initial return IPOs (group 1) and 3.53% for the largest initial return IPOs (group 4). The JonckheereTerpstra test supports the increasing trend in turnover ratio, with a JTstatistic of 4.67. The increasing trend remains significant in the other three price groups (Panels A2 to A4). We also repeat this analysis in the first 20tradingday period. The associations between initial return and liquidity measures do not have significant changes.
effects of these events, we divide our IPO sample into two groups: preOHR/16th and postOHR/16th and compare the spread measures across initial return groups in each group. All spread measures declined significantly after the implementation of OHR and tick size reduction in 1997, but the increasing/ decreasing patterns across the initial return groups were essentially unchanged. These results are available from the authors upon request.
11
As an additional test, we compare the percentage spread and turnover ratio by controlling for trading volume. To do this, we first divide the sample IPOs into fourbyfour groups based on initial return and trade volume. Then, we compare the percentage spread and turnover ratio across initial return groups within each of the four volume groups. We conduct these tests in both the first 20tradingday period and in the 240tradingday period. The results (not reported) are similar to the results reported in Table 2. In summary, the univariate analyses demonstrate that initial return is positively related to aftermarket liquidity. Additional evidence and a detailed discussion of this are provided in Section 5.
4. Univariate analyses on the relation between share retention and liquidity The sharefloating effect on liquidity proposed by Zheng, Ogden, and Jen (2003) suggests that high retention by the preIPO owners would reduce the secondarymarket liquidity, since high retention per se reduces the number of shares floating in the market. In contrast, the signaling theory implies that high retention would increase market liquidity, since the “good signal” of high retention attracts more trades. As a preliminary test, we compare percentage spread and turnover ratio across retention groups. We divide the whole sample into four groups based on retention rate. Group 1 (4) includes IPOs with the lowest (highest) retention rate. The average retention rate ranges from 46% to 84%. Table 4 reports the mean values of the two liquidity measures for each retention group. The percentage spread is negatively related to retention rate. The results are significant in both the first 20tradingday period and in the 240tradingday period. For example, in the first 20tradingday period (Panel A), the percentage spread is 3.17% for the IPOs with the lowest retention (group 1), compared with 2.15% for those with the highest retention (group 4). The JonckheereTerpstra test supports the decreasing trend, with a JTstatistic of 11.49. The turnover ratio is positively related to the retention rate. It increases from 7.18% for retention group 1 to 13.22% for retention group 4 in the first 20tradingday period, with a JT Z4statistic of 12.21. The
12
positive relation remains significant in the 240tradingday period. These results suggest that retention is positively related to liquidity, consistent with the signaling theory that high retention sends out a good signal to investors and attracts more trades in the aftermarket.
5. Robustness test 5.1 Multivariate regressions The microstructure literature provides evidence that firm size, trade volume, price, risk, and insiders’ trading are the main determinants of spreads (Huang and Stoll, 1997; Stoll, 2000). To see how initial return and share retention affect spreads and turnover ratio after controlling for these factors, we run the following regression: Liquidity = b0 + b1RT (or IR) + b2LogPrice + b3LogVol + b4LogNSO + b5LogRisk + b6ISDT + b7Tick + µ .
(1)
Where, Liquidity is measured by Percentage Spread (quoted spread/quoted midpoint) and Turnover (daily trading volume/Total number of shares offered). We repeat the above regression with each of the liquidity measure. The regressors are defined as: RT (retention rate) = the number of shares retained by preIPO shareholders/total number of shares outstanding after the IPO; IR (initial return) = 100x(Close – Offer)/Offer, where Close and Offer are the first day closing price and offer price, respectively; Price = the daily average of quoted midpoint at bidask prices in the aftermarket; Risk = the standard deviation of daily return based on closing price during the regression periods (20 trading days and 240 trading days); Vol = the average daily number of shares traded;
13
NSO = the number of shares outstanding in the regression period (20 trading days and 240 trading days), a proxy for firm size. 7 ISDT = the average daily insiders’ trading volume / total daily trading volume; and Tick = 1 for the IPOs after the tick size reduction in June 1997 and 0 otherwise. Bradley and Jordan (2002) find that initial return and retention rate are highly correlated. To avoid a potential multicolinearity problem between these two variables, we add initial return and retention rate separately in the liquidity regression (eq.1). Table 5, Panel A reports the regression results excluding retention rate during the first 20tradingday period. Note that price is not included in the percentage spread regression, since percentage spread already controls for the price effect. Similarly, trading volume is not included in the turnover ratio regression. In the percentage spread regression, initial return (IR) is negatively related to percentage spreads with a coefficient of –0.245. The negative relation is significant after controlling for other variables, with a tstatistic of 3.33. Other regression results are also worthy of mention. The coefficients on trading volume, the number of shares outstanding (NSO), and the dummy variable Tick are negative and significantly different from zero. These results suggest that large firms and actively traded stocks have smaller spreads (high liquidity), consistent with the findings of Kavajecz (1999) and Lakonishok and Lee (2001). The negative coefficient on tick size (Tick) confirms the smaller spreads after tick size reduction. The positive relation between percentage spread and return volatility (risk) suggests that market makers require large spreads for trading
7
We also use other proxies for firm size such as total assets value and market value of equity. The
regression results are qualitatively similar. However, we prefer using the number of shares outstanding to measure firm size for two reasons. First, unlike the market value of equity, this measure is relatively stable and is not affected by stock price. Second, the data for total assets from Compustat are reported either quarterly or annually. The time frame for this data does not match exactly with our test periods, whereas the daily data for the number of shares outstanding from CRSP have exact matches.
14
risky stocks. Similar to the findings of Cao, Field, and Hanka (2004), the results indicate that the insider trading does not have a significant effect on spreads. In the turnover ratio regression, initial return is positively related to turnover ratio with a coefficient of 4.393. The positive relation is significant after controlling for other variables, with a tstatistic of 10.05. The coefficients on price, risk, and tick size are also positive and significantly different from zero. The positive relation between risk and spreads and turnover suggests that the riskier a stock is, the larger the spreads and the higher the turnover. This is consistent with Carter and Manaster’s (1990) argument that if investors have scarce resources to invest in information acquisition, they specialize in acquiring information for the most uncertain investments. The positive relation between risk and turnover supports the view that the difference in investors’ opinions attracts trades. The positive coefficient on tick size indicates more active trading after tick size reduction. We replace initial return with retention rate and repeat the above regression. The results are reported in Table 5, Panel B. Although the correlation between percentage spread and retention rate is negative, it is not significant. The positive relation between retention rate and turnover ratio is significant after controlling for other variables. The regression coefficients on other variables are similar to those in the initial return regression. As an additional test, we rerun the above regressions during a 240tradingday period and investigate whether the results are sensitive to the test period. Table 5, Panel C reports the regression results excluding retention rate. After controlling for other variables, initial return is positively related to turnover ratio and negatively related to the percentage return. These results are consistent with those in the 20tradingday period. The coefficients on other variables are similar to those in the 20tradingday period. The regression results on retention rate are presented in Table 5, Panel D. The retention rate is positively related to the turnover ratio and negatively related to the percentage spread.
15
In summary, both the univariate and the regression analyses indicate that the effects of initial return and retention rate on percentage spread and turnover are not sensitive to test periods. The percentage spread is negatively correlated with both initial return and retention rate, whereas the turnover ratio is positively related to initial return and retention rate. These results suggest that initial return and retention rate are positively related to liquidity in the IPO aftermarket.
5.2 Instrumental variable approach and correction of endogeneity problem The regression results showing the negative (positive) relation between percentage spread (turnover ratio) and initial return may be biased. Bias could be caused by an endogeneity problem for initial return and retention. The literature suggests that initial return is not exogenous and is affected by many factors. For example, Rock (1986) notes that underwriters and issuers use underpricing to attract and reward uninformed investors since these investors face a “winner’s curse” problem. Baron (1982) and Habib and Ljungqvist (2001) argue that underwriters use underpricing as a substitute for promotion and marketing expenses. We use the instrumental variable approach to correct for the endogeneity problem. Specifically, we first regress initial return (IR) and retention rate (RT) separately on a set of instrumental variables, which are directly associated with these variables, but have zero or low correlation with other variables in the regression of interest. Then, we use the fitted values of initial return and retention rate in the liquidity regressions (eq.1). Dummy variables comprise the first set of instrumental variables. Ljungqvist and Wilhelm (2003) find that the initial return during the bubble period (1999 and 2000) is significantly larger than in other years.8 Megginson and Weiss (1991), Schultz (1993), Brav and Gompers (2003), and Bradley and Jordan (2002) note that the firstday return also depends on whether or not an IPO is backed by a venture capitalist (VC). According to Loughran and Ritter
8
We find similar results as reported in Table 1.
16
(2004) and Chemmanur and Yan (2003), IPOs in hightechnology industries exhibit larger initial returns. Thus, we use Bubble as a dummy variable for the Internet bubble period, VC for IPOs backed by VCs, and Tech for IPOs in hightechnology industries. 9 The second group of instrumental variables includes underwriters’ reputation rank (Rank). An underwriter’s prestige affects underpricing because of the certification role of underwriter’s reputation (Carter, Dark, and Singh, 1998; Chen and Mohan, 2002).10 The number of shares outstanding (NSO) after the offering is a proxy for firm size, since risk is lower and initial return is smaller for large and established firms (Beatty and Ritter, 1986; Carter, Dark, and Singh, 1998). Theory (Brennan and Franks, 1997; Stoughton and Zechner, 1998) also posits that initial return is used to achieve desired ownership structure. Hence, we also include the number of shareholders (NSH) after the offerings. We collect the total number of shareholders from Compustat (data item #100). Since this data is reported only in the Compustat annual data file, we use the number of shareholders in the year that an IPO came to the market. The last group of instrumental variables is related to the public information in the market and the information revealed during the bookbuilding process. The first variable in this set is price adjustment (Diff), the difference between the final offer price and the midpoint of initial filing range. Hanley (1993) and Bradley and Jordan (2002) find that the adjustment of offer price relative to the initial filing range reveals investors’ private information level and is positively related to initial return. The other variables include the average returns of IPOs in the past month (Lag) and the market return before an IPO came to the market (Cum). The literature indicates that these variables have a predictive power on initial return. 9
As in Loughran and Ritter (2004), Ljungqvist and Wilhelm (2003), and Chemmanur and Yan (2003),
hightechnology industries are those with the following SIC codes: 3571, 3572, 3575, 3577, 3578, 3661, 3663, 3669, 3674, 3812, 3823, 3825, 3826, 3827, 3829, 4899, 7370, 7371, 7372, 7374, 7375, 7378, and 7379. 10
We collect underwriters’ reputation ranks from Jay Ritter's web: http://bear.cba.ufl.edu/ritter/rank.pdf.
17
Similarly, the retention rate is regressed on the instrumental variables listed above. The regression is summarized as the follows: IR (or RT) = a0 + a1RT (or IR) + a2LogNSH +a3Log (NSO) + a4VC + a5Rank + a6Diff + a7Lag+ a8Cum+ a9Tech + a10Bubble + ε .
(2)
where IR and RT are the initial return and retention rate, respectively, as defined previously. Other variables are defined as: LogNSH = log of number of shareholders; Log(NSO) = log of number of shares outstanding after the IPO; VC= 1 for IPOs backed by venture capitalists and 0 otherwise; Rank = underwriter’s reputation rank, takes a value between one and ten; Diff = the difference between offer price and the midpoint of initial filing range; Lag = the average initial return of all IPOs in the past month; Cum = cumulative return on the NASDAQ index 15 trading days before the issue date; Tech = 1 for hightechnology industries and 0 otherwise; and Bubble = 1 for IPOs in 1999 or 2000 and 0 otherwise. For expositional purposes, Table 6, Panel A reports the results of the first stage regressions (eq.2). Consistent with the literature, initial return and retention rate are positively correlated. Initial return is also positively related to the number of shares outstanding, the offer price adjustment (Diff), market return, and the bubble period. The effects of other variables on initial return are not significant. Retention rate is positively related to both the number of shares outstanding and to high technology industries, similar to the findings of Ljungqvist and Wilhelm (2003). The large Rsquares, especially on the initial return regression (AdjR2 = 42.4%), imply that the instrumental variables provide a good explanation of initial return and retention rate. Table 6, Panels B and C report the second stage regression (eq.1) results with the fitted values of IR and RT in the first 20tradingday period. Consistent with the results in Tables 2 and 5, both initial return and retention rate are negatively related to percentage spread, but positively
18
related to turnover ratio after controlling for other factors. The coefficients on other variables are similar to those reported in Table 5. We repeat this regression using the data during a 240tradingday period and report the results in Panels D and E of Table 6. The results are similar to those in the first 20tradingday period.
5.3 Factor analysis by principal components As an additional robustness test, we conduct factor analysis by principal components. The correlation coefficients between percentage spread, turnover ratio, initial return, and retention rate are significant. For instance, the coefficient between initial return and percentage spread is

0.3187, and the coefficient between initial return and turnover ratio is 0.5145. Similarly, the coefficient between retention rate and percentage spread (turnover ratio) is –0.1963 (0.2939). These again confirm the results that the effects of initial return and retention rate on percentage spread and turnover ratio are significant.
6. Conclusions Although liquidity theory suggests that underpricing improves aftermarket liquidity, limited evidence exists. The impact of share retention on aftermarket liquidity is far less clear. A high retention rate, per se, reduces the number of shares floating in the market and could have a negative effect on market liquidity. On the other hand, a high retention rate signals positive information about a firm and could attract aftermarket trades and improve market liquidity. We investigate these two issues and have the following main findings: First, the proportion of shares held by the preIPO owners is positively related to turnover ratio and negatively related to percentage spreads. The relations are significant after controlling for other variables. This suggests that the signaling effect of share retention dominates the effect of share floatation. A higher retention rate attracts more trades in the aftermarket and improves market liquidity.
19
Second, we find that initial return is positively related to turnover and negatively related to percentage spread. The relations are robust and stable over different periods. These findings suggest that initial return is positively correlated with liquidity, supporting Booth and Chua’s (1986) liquidity theory. Consistent with the literature, our results provide additional evidence that firm size, trade volume, and share price have significant effects on aftermarket liquidity. Large firms and actively traded stocks have smaller spreads, larger turnover ratio, and higher liquidity.
20
REFERENCES Aggarwal, R. (2000) “Stabilization activities by underwriters after initial public offerings”, The Journal of Finance, Vol. 55, No.3, pp.10751103. Baron, D.P. (1982) “A model of the demand for investment banking advising and distribution services for new issues”, The Journal of Finance, Vol. 37 No4, pp.955976. Bayley, L., P. Lee, and T. Walter (2003) “IPO flipping in Australia: crosssectional explanations”, Working paper, the University of New South Wales. Beatty, R. and J. Ritter (1986) “Investment banking, reputation and the underpricing of initial public offerings”, Journal of Financial Economics, Vol. 15 No.12, pp.213232. Benveniste, L. and P. Spindt (1989) “How investment bankers determine the offer price and allocation of new issues”, Journal of Financial Economics, Vol. 24 No.2, pp.343362. Bhushan, R. (1989) “Firm characteristics and analyst following”, Journal of Accounting and Economics, Vol. 11 No23, pp.255274. Booth, J. and L. Chua (1996) “Ownership dispersion, costly information, and IPO underpricing”, Journal of Financial Economics, Vol. 41 No. 2, pp.291310. Bradley, D. and B. Jordan (2002) “Partial adjustment to public information and IPO underpricing”, Journal of Financial and Quantitative Analysis, Vol. 37 No.3, pp.595616. Brav, A. and P. Gompers (2003) “The role of lockups in initial public offerings”, The Review of Financial Studies, Vol.16 No.1, pp.129. Brennan, M. and J. Franks (1997) “Underpricing, ownership, and control in initial public offerings of equity securities in the UK”, Journal of Financial Economics, Vol.45 No.3, pp.391413. Cao, C., L. Field, and G. Hanka (2004) “Does insider trading impair market liquidity? Evidence from IPO lockup expiration”, Journal of Financial and Quantitative Analysis, Vol.39 No.4, pp.2546.
21
Carter, R., F. Dark, and A. Singh (1998) “Underwriter reputation, initial returns, and the longrun underperformance of IPO stocks”, The Journal of Finance, Vol.53 No.1, pp.285311. Carter, R. and S. Manaster (1990) “Initial public offerings and underwriter reputation”, The Journal of Finance, Vol.45 No.4, pp.10451067. Chemmanur, T. and A. Yan (2003) “Product market advertising and initial public offerings: Theory and empirical evidence”, Working paper, Boston College. Chen, C. and N. Mohan (2002) “Underwriter spread, underwriter reputation, and IPO underpricing: A simultaneous equation analysis”, Journal of Business Finance and Accounting, Vol.29 No. 34, pp.521540. Christie, W. and P Schultz (1994) “Why do NASDAQ market makers avoid oddeighth quotes?” The Journal of Finance, Vol.49 No.5, pp.18131840. Demsetz, H., (1968) “The cost of transacting”, Quarterly Journal of Economics, Vol. 83, pp.33 53. Downes, D. and R. Heinkel (1982) “Signaling and the valuation of unseasoned new issues”, The Journal of Finance, Vol.37 No.1, pp.110. Ellis, K., R. Michaely, and M. O’Hara (2000) “When the underwriter is the market maker: An examination of trading in the IPO aftermarket”, The Journal of Finance, Vol. 55 No.3, pp.10391074. Field, L. and D. Sheehan (2004) “IPO underpricing and outside blockholdings”, Journal of Corporate Finance, Vol. 10 No.2, pp.263280. Habib, M. and A. Ljungqvist (2001) “Underpricing and entrepreneurial wealth losses in IPOs: Theory and evidence”, The Review of Financial Studies, Vol.14 No. 2, pp.433458.
Hanley, K. (1993) “The underpricing of initial public offerings and the partial adjustment phenomenon”, Journal of Financial Economics, Vol. 34, pp.231250.
22
Hegde, S. and R. Miller (1989) “Marketmaking in initial public offerings of common stocks: An empirical analysis”, Journal of Financial and Quantitative Analysis, Vol.24 No.1, pp.7590. Huang, R. and H. Stoll (1997) “The components of the bidask spread: a general approach”, Review of Financial Studies, Vol.10 No.4, pp.9951034. Jain, B. and O. Kini (1994) “The postissue operating performance of IPO firms”, The Journal of Finance, Vol.49 No.5, pp.16991726. Kavajecz, K. (1999) “A specialist’s quoted depth and the limit order book”, Journal of Finance, Vol.54 No.2, pp.747771. Koh, F. and T. Walter (1989) “A direct test of Rock’s model of the pricing of unseasoned issues”, Journal of Financial Economics, Vol.23 No.2, pp.251272. Lakonishok, J. and I. Lee (2001) “Are insider trades informative”? Review of Financial Studies, Vol.14 No.1, pp.79111. Lee, P., S. Taylor, and T. Walter (1996) “Australian IPO pricing in the short and long run”, Journal of Banking and Finance, Vol.20 No.7, pp.11891210. Leland, H. and D. Pyle (1977) “Information asymmetries, financial structure, and financial intermediation”, The Journal of Finance, Vol.32 No.2, pp.371387. Ljungqvist, A. and W. Wilhelm (2003) “IPO pricing in the Dotcom bubble”, The Journal of Finance, Vol.58 No.2, pp.723752. Loughran, T. and J. Ritter (2004) “Why has IPO underpricing changed over time”? Financial Management, Vol.33 No., pp.537. McInish, T. and R. Wood (1992) “An analysis of intraday patterns in bid/ask spreads for NYSE stocks”, The Journal of Finance, Vol.47 No.2, pp.753764. Megginson, W. and K. Weiss (1991) “Venture capitalist certification in initial public offerings”, The Journal of Finance, Vol.46 No.3, pp.879903. Merton, R. (1987) “A simple model of capital market equilibrium with incomplete information”, The Journal of Finance, Vol.42 No.3, pp.483510.
23
Pham, P., P. Kalev, and A. Steen (2003) “Underpricing, stock allocation, ownership structure and postlisting liquidity of newly listed firms”, Journal of Banking and Finance, Vol.27 No.5, pp.919947. Ritter, J. and I. Welch (2002) “A review of IPO activities, pricing, and allocations”, The Journal of Finance, Vol.57 No.4, pp.17951828. Rock, K. (1986) “Why new issues are underpriced”, Journal of Financial Economics, Vol.15 No.12, pp.187212. Schultz, P. (1993) “Unit initial public offering”, Journal of Financial Economics, Vol.34 No.2, pp.199229. Schultz, P. and M. Zaman (2001) “Do the individuals closest to Internet firms believe they are overvalued”? Journal of Financial Economics, Vol.59 No.3, pp.347381. Stoll, H. (2000) “Friction”, The Journal of Finance, Vol.55 No.4, pp.14791514. Stoughton, N. and J. Zechner (1998) “IPOmechanisms, monitoring and ownership structure”, Journal of Financial Economics, Vol.49 No.1, pp.4577. Zheng, S., J. Ogden, and F. Jen (2003) “Pursuing value through liquidity: Share retention, lockup, and underpricing in IPOs”, Working paper, University of Manitoba.
24
Table 1 Descriptive statistics of initial public offerings The IPOs in the test sample include 1,673 common stocks listed on NASDAQ with offer prices equal to or greater than $5 during the period from 1993 to 2000. Initial return is defined as 100x(Close – Offer)/Offer, where Close and Offer are the first day closing price and offer price, respectively. Retention rate is defined as the ratio of the number of shares retained by preIPO shareholders to total number of shares outstanding after the IPO. We report the mean (median) values of the related variables. Year
N
Offer Price ($)
Initial Return (%) 14.98 (8.33)
Shares Offered (000 unit) 2,328 (2,050)
Retention Rate 0.65 (0.67)
1993
243
12.00 (12.00)
1994
195
10.36 (10.00)
11.41 (5.49)
2,058 (2,000)
0.63 (0.67)
1995
248
12.04 (12.00)
23.01 (16.67)
2,574 (2,400)
0.66 (0.68)
1996
333
12.23 (12.00)
17.48 (11.90)
2,952 (2,500)
0.68 (0.71)
1997
221
11.38 (11.00)
14.98 (10.29)
3,158 (2,750)
0.64 (0.70)
1998
96
11.53 (12.00)
24.23 (8.33)
3,030 (2,800)
0.68 (0.72)
1999
192
12.77 (12.00)
52.54 (30.12)
4,329 (4,000)
0.75 (0.79)
2000
145
13.16 (12.00)
43.98 (22.92)
5,334 (4,750)
0.77 (0.79)
Whole Period
1673
11.94 (12.00)
23.54 (11.76)
3,090 (2,583)
0.68 (0.71)
25
Table 2 Relation between initial return and aftermarket liquidity We divide our IPO samples into four groups based on the initial return (IR). Group 1 (4) includes IPOs with the smallest (largest) IR. IR = 100x(Close – Offer)/Offer, where Close and Offer are the first day closing price and offer price, respectively. We use both percentage spread and turnover ratio to measure liquidity. Percentage Spread is defined as 100xSpread/Midpoint, where spread is the quoted bidask spread (i.e., the difference between ask and bid prices), and midpoint is the average of quoted bidask prices. Turnover is calculated as 100xdaily trading volume (shares)/number of shares offered. For expositional purpose, we also report Price the average quoted midpoint at bidask prices in the IPO aftermarket. Both the percentage spread and price are the intraday averages during a 20tradingday period (Panel A1 and Panel B1) and a 240tradingday period (Panel A2 and Panel B2) after the initial offerings. We conduct JonckheereTerpstra (JT) nonparametric test (onetailed analysis) to analyze the increasing/decreasing trend across the four initial return groups. The plus/minus signs of the JT Zstatistic capture the increasing/decreasing trend of the test variables. Panel A: Whole Sample (IPOs from 1993 to 2000) Whole Group Group 1 Group 2 Group 3 N = 1760 N = 417 N = 418 N = 419 IR = 23.55% IR= 2.24% IR= 6.15% IR=19.46% Panel A1: Group averages during the first 20tradingday period
Group 4 N = 416 IR=71.06%
JonckheereTerpstra Zstatistic
Percentage Spread ($)
2.88
3.22
3.36
2.89
2.06
13.09***
Turnover (%)
9.38
6.85
6.49
9.11
15.10
21.42***
Price ($)
15.72
10.49
11.83
15.52
25.08
26.97***
Panel A2: Group averages during a 240tradingday period Percentage Spread ($)
3.30
3.77
3.72
3.19
2.51
13.05***
Turnover (%)
3.70
2.81
2.73
3.47
5.82
16.80***
Price
16.37
11.84
13.23
16.91
23.50
18.00***
26
Table 2 (continued) Panel B: Subsample, excluding IPOs in 1999 and 2000 Whole Group Group 1 Group 2 Group 3 N = 1336 N = 326 N = 384 N = 369 IR = 17.16% IR= 1.09% IR= 6.15% IR=19.27% Panel B1: Group averages during the first 20tradingday period
Group 4 N = 257 IR=69.67%
JonckheereTerpstra Zstatistic
Percentage Spread ($)
3.26
3.65
3.51
3.09
2.63
10.17***
Turnover (%)
7.53
5.53
6.08
8.27
11.19
19.11***
Price ($)
14.28
10.51
11.81
15.31
21.28
21.78***
Panel B2: Group averages during a 240tradingday period Percentage Spread ($)
3.34
3.96
3.75
3.33
2.97
8.72***
Turnover (%)
3.25
2.61
2.73
3.19
4.90
13.23***
Price
15.75
12.27
13.56
16.82
21.91
14.96***
*** indicates the 0.01 significance level.
27
Table 3: Comparing liquidity by controlling for share price We divide sample IPOs into 4 x 4 groups based on initial return (underpricing) and aftermarket share price during a 240tradingday period in the aftermarket. For initial return classification, group 1 (4) includes IPOs with the smallest (largest) initial returns. Similarly, for price classification, group 1 (4) includes IPOs with the lowest (highest) price in the IPO aftermarket. Initial return is calculated as 100x(Close – Offer)/Offer, where Close and Offer are the first day closing price and offer price, respectively. We use both percentage spread and turnover ratio to measure liquidity. Percentage Spread is defined as 100xSpread/Midpoint, where spread is the quoted bidask spread (i.e., the difference between ask and bid prices), and midpoint is the average of quoted bidask prices. Turnover is calculated as 100xdaily trading volume (shares)/number of shares offered. We compare the liquidity measures across initial return groups by controlling for price. JonckheereTerpstra (JT) nonparametric test (onetailed analysis) is used to analyze the increasing/decreasing trend across groups. The plus/minus signs of the JT Zstatistic capture the increasing/decreasing trend of the test variables. Initial Return Groups 1
Percentage Turnover (%) Spread (%) A1. Price Group = 1 4.75 2.56
Percentage Turnover (%) Spread (%) A2. Price Group = 2 3.55 2.66
2
4.61
2.46
3.90
2.40
3
4.55
2.82
3.75
2.89
4
4.57
3.53
3.24
4.18
JT Zstatistic
4.67***
1
0.50 A3. Price Group = 3 2.85
2.90
0.41 5.42*** A4. Price Group = 4 2.14 3.86
2
3.09
3.11
2.35
3.52
3
2.87
3.43
2.29
4.34
4
2.61
5.04
1.75
7.22
JT Zstatistic
2.53***
7.46***
5.82***
7.38***
*** indicates the 0.01 significance level.
28
Table 4 Relation between retention rate and aftermarket liquidity We divide IPO samples into four groups based on retention rate (RT). Group 1 (4) includes IPOs with the lowest (highest) RT. RT is the ratio of the number of shares retained by preIPO shareholders to total shares outstanding after the IPO. We use both percentage spread and turnover ratio to measure liquidity. Percentage Spread is defined as 100xSpread/Midpoint, where spread is the quoted bidask spread (i.e., the difference between ask and bid prices), and midpoint is the average of quoted bidask prices. Turnover is calculated as 100xdaily trading volume (shares)/number of shares offered. Both the percentage spread and prices are the intraday averages during a 20tradingday period (Panel A) and a 240tradingday period (Panel B) after the initial offerings. We conduct JonckheereTerpstra (JT) nonparametric test (onetailed analysis) to analyze the increasing/decreasing trend across the four initial return groups. The plus/minus signs of the JT Zstatistic capture the increasing/decreasing trend of the test variables. Group 1 Group 2 Group 3 Group 4 JonckheereTerpstra N = 414 N = 422 N = 418 N = 419 Zstatistic RT = 0.46 RT = 0.66 RT = 0.74 RT= 0.84 Panel A: Group averages during the first 20tradingday period Percentage Spread (%) 3.17 3.26 2.93 2.15 11.49*** Turnover (%) 7.18 7.75 9.39 13.22 12.21*** Panel B: Group averages during the whole period (240 trading days) Percentage Spread (%) 3.72 3.68 3.34 2.46 12.79*** Turnover (%) 2.68 3.01 3.69 *** indicates the 0.01 significance level, respectively.
29
5.43
13.09***
Table 5 Regression of liquidity This table reports the results of the following regression: Liquidity = b0 + b1RT (or IR) + b2LogPrice + b3LogVol + b4LogNSO + b5LogRisk + b6ISDT + b7 Tick + µ .
(1)
Where, liquidity is measured by percentage spread and turnover ratio. Percentage spread is defined as 100xSpread/Midpoint, where spread is the quoted bidask spread (i.e., the difference between ask and bid prices), and midpoint is the average of quoted bidask prices. Turnover is calculated as 100xdaily trading volume (shares)/number of shares offered. RT is retention rate, the number of shares retained by preIPO shareholders divided by the total number of shares outstanding after the IPO. IR is initial (first day) return calculated as 100x(Close – Offer)/Offer, where Close and Offer are the firstday closing price and offer price, respectively. Price is the daily average of quoted midpoint at bidask prices in the aftermarket. Risk is the standard deviation of daily return based on closing price during the regression periods (20 trading days and 240 trading days). Vol is daily number of shares traded. NSO is the number of shares outstanding during the regression period. ISDT is the ratio of daily insiders’ trading volume to the total trading volume. Tick = 1 for the IPOs after the tick size reduction in June 1997 and 0 otherwise. Numbers in parentheses are regression tstatistics.
30
Table 5 (continued) LHS Intercept Variables Percentage Spread
18.97*** (32.88)
Turnover
10.76*** (4.53)
Percentage Spread
19.17*** (33.24)
Turnover
9.91*** (4.06)
Percentage Spread
29.79*** (26.27)
Turnover
5.09*** (4.47)
Percentage Spread
10.26*** (47.42)
Turnover
5.601*** (4.81)
RT
IR
LogPrice
LogVol
LogNSO
LogRisk
ISDT
Panel A: Regression during the first 20tradingday period, excluding retention rate (RT) 0.245*** 0.956*** 0.155*** 0.789*** 0.008 (3.33) (21.97) (2.88) (13.28) (0.62) 4.393*** 3.460*** 0.349 4.848*** 0.009 (10.05) (9.44) (1.57) (17.97) (1.46) Panel B: Regression during the first 20tradingday period, excluding initial return (IR) 0.045 0.994*** 0.138** 0.767*** 0.001 (0.26) (23.13) (2.35) (12.73) (0.50) 2.477*** 5.716*** 0.104 5.390*** 0.013** (2.93) (19.32) (0.43) (19.88) (1.92) Panel C: Regression during the whole period (240 trading days), excluding retention rate (RT) 0.315*** (4.23)
1.528*** (30.22)
0.132** (2.35)
2.487*** (28.42)
0.003 (1.55)
1.789*** 1.967*** 0.230** 2.959*** 0.003 (10.39) (16.33) (2.36) (17.03) (0.76) Panel D: Regression during the whole period (240 trading days), excluding initial return (IR) 0.588*** 1.610*** 0.039 2.504*** 0.003 (3.38) (33.05) (0.65) (28.37) (1.54) 1.955*** (5.13)
2.499*** (22.74)
0.057 (0.53)
***, **, and * indicate the 0.01, 0.05, and 0.1 significance level, respectively.
31
3.390*** (19.78)
0.005 (1.03)
Tick
AdjR2
1.226*** (19.65)
0.576
2.707*** (9.08)
0.526
1.242*** (19.90)
0.573
3.176*** (10.47)
0.501
0.966*** (15.09)
0.594
0.004 (0.73)
0.456
1.011*** (15.83)
0.592
0.263* (1.86)
0.432
Table 6 Instrumental regression This table reports the results of the following regression: Liquidity = b0 + b1RT (or IR) + b2LogPrice + b3LogVol + b4LogNSO + b5LogRisk + b6ISDT + b7 Tick + µ , IR (or RT) = a0 + a1RT (or IR) + a2LogNSH +a3Log(NSO) + a4VC + a5Rank + a6Diff + a7Lag+ a8Cum+ a9Tech + a10Bubble + ε .
(1) (2)
Where, liquidity is measured by percentage spread and turnover ratio. Percentage spread is defined as 100xSpread/Midpoint, where spread is the quoted bidask spread (i.e., the difference between ask and bid prices), and midpoint is the average of quoted bidask prices. Turnover is calculated as 100xdaily trading volume (shares)/number of shares offered. IR is initial (first day) return calculated as 100x(Close – Offer)/Offer, where Close and Offer are the firstday closing price and offer price, respectively. RT is retention rate, the number of shares retained by preIPO shareholders divided by total shares outstanding after the IPO. Price is the daily average of quoted midpoint at bidask prices in the aftermarket. Risk is the standard deviation of daily return based on closing price during the regression periods (20 trading days and 240 trading days). Vol is daily number of shares traded. NSO is the number of shares outstanding during the regression period. ISDT is the ratio of daily insiders’ trading volume to the total trading volume. Tick = 1 for the IPOs after the tick size reduction in June 1997 and 0 otherwise. NSH is the number of shareholders during the regression period. VC = 1 for IPOs backed by venture capitalists and 0 otherwise. Rank is underwriter’s reputation rank taking a value from 1 to 10. Diff is the difference between the offer price and the midpoint of initial filing range. Lag is the average daily initial return for all IPOs in the recent past month. Cum is the cumulative return on the NASDAQ index 15 trading days before the issue date. Tech = 1 for IPOs in hightechnology industries and 0 otherwise. Bubble = 1 for IPOs in 1999 or 2000 and 0 otherwise. We first run IR and RT regressions (eq. 2) and use the fitted values of IR and RT in liquidity regressions (eq. 1). Numbers in parentheses are regression tstatistics.
LHS Variable IR
IR
RT
0.042*** (2.69)
Panel A: The first stage regression LogNSO VC Rank Diff
RT
LogNSH
0.219*** (2.69)
0.010 (1.26)
0.054** (2.24)
0.032 (1.26)
0.001 (0.13)
0.002 (0.66)
0.109*** (11.10)
0.005 (0.43)
0.006* (1.90)
32
Lag
Cum
Tech
Bubble
AdjR2
0.967*** (16.74)
0.106 (1.32)
1.190** (2.17)
0.034 (1.18)
0.143*** (2.95)
0.424
0.059** (2.00)
0.029 (0.82)
0.051 (0.21)
0.035*** (2.75)
0.023 (1.06)
0.264
Table 6 (continued) LHS Variables Percentage Spread Turnover
Percentage Spread Turnover
LHS Variables Percentage Spread Turnover
Percentage Spread Turnover
Panel B: Second stage regression in the first 20tradingday period, excluding retention rate (RT) Fitted Fitted LogPrice LogVol LogNSO LogRisk ISDT Tick RT IR 18.33*** 0.549*** 0.964*** 0.030 0.872*** 0.001 1.308*** (20.47) (2.99) (14.70) (0.36) (9.49) (0.76) (13.85) Intercept
10.10*** (2.62)
AdjR2 0.602
2.102* 5.210*** 0.120 5.282*** 0.011 2.738*** (1.87) (9.39) (0.35) (12.38) (1.14) (5.84) Panel C: Second stage regression in the first 20tradingday period, excluding initial return (IR) 19.07*** 0.297** 1.025*** 0.081 0.832*** 0.001** 1.352*** (20.20) (2.19) (16.35) (0.41) (8.80) (0.64) (14.38)
0.534
15.40*** 30.65*** 5.700*** 2.927*** 5.068*** 0.011 2.871*** (3.68) (3.66) (13.52) (3.16) (11.86) (1.10) (6.35) Panel D: Second stage regression in the whole period (240 trading days), excluding retention rate (RT) Intercept Fitted Fitted LogPrice LogVol LogNSO LogRisk ISDT Tick RT IR 13.34*** 0.349** 1.636*** 0.016 2.583*** 0.015** 1.125*** (19.19) (2.03) (22.97) (0.19) (20.15) (2.14) (12.15)
0.540
0.637
AdjR2 0.636
6.827*** 0.789** 2.516*** 0.001 3.238*** 0.031** 0.121 (4.17) (2.08) (16.22) (0.09) (13.26) (2.08) (0.62) Panel E: Second stage regression in the whole period (240 trading days), excluding initial return (IR) 30.15*** 2.602** 1.661*** 0.227 2.597*** 0.016** 1.160*** (30.14) (2.19) (23.83) (1.19) (19.87) (2.24) (12.74)
0.491
10.56*** (6.05)
0.509
20.46*** (5.86)
2.560*** (18.14)
2.033*** (5.30)
***, **, and * indicate the 0.01, 0.05, and 0.1 significance level, respectively.
33
2.963*** (12.30)
0.028* (1.93)
0.202 (1.08)
0.635