Venture capital and IPO lockup Expiration: An Empirical Analysis.

Abstract

Most initial public offerings (IPOs) feature "lockup" agreements, which bar insiders from selling the stock for a set period following the IPO, usually 180 days. We examine stock price behavior in the period surrounding lockup expiration for a sample of 2,529 firms from 1988

to 1997. We find that lockup expirations are, on average, associated with significant and negative abnormal returns, but the losses are concentrated in firms with venture capital backing. For the venture-capital-backed group, the largest losses occur for high-tech firms and firms with the greatest post-IPO stock price increases, the largest relative trading volume in the period surrounding expiration, and the highest quality underwriters.

I. Introduction

Initial public offerings (IPOs) generally feature "lockup" agreements under which corporate insiders are prohibited from selling shares before a certain date, ranging from a month to several years after the IPO. Once the lockup period is over, insiders are free to sell, although they remain subject to more general insider trading regulations.

In both theory and practice, insider sales play an important role because of potential informational asymmetries. Because the lockup expiration date represents the first opportunity for insiders to sell, significant share price revisions are possible as market participants infer private information from insider activity. In addition, the percentage of shares subject to lockup often exceeds 100 percent of the shares in public hands. As a result, the possibility of a large, sudden increase in supply exists, which may also affect share value.

Interest in lockup expirations has grown, as evidenced by at least four recent working papers (Brau (1999), Bray and Gompers (1999), Field and Hanka (2001), Ofek and Richardson (2000)), a Web site (www.ipolockup.com), and numerous articles in the popular press, including, as of February 14, 2000, a regular feature in the Wall Street Journal. Our goal in this article is to add to this literature by providing a detailed analysis of share price reactions to IPO lockup expirations. Based on a sample of 2,529 IPOs for the ten-year period ending in 1997, we find that lockup expirations are associated with significant price declines. The average abnormal return on the lockup expiration day is -.74 percent, and the cumulative abnormal return over the five-day surrounding period is -1.61 percent, both of which are significant at conventional levels. Moreover, the cumulative loss does not appear to be a transitory effect.

On closer inspection, the negative abnormal returns in the period surrounding lockup expirations are largely concentrated in the 45 percent of the firms in our sample with venture capital (VC) backing. Such firms lose, on average, 3 percent to 4 percent of their value in this period, and high-tech firms with VC backing are particularly hard hit. Non-VC-backed firms lose relatively little value, regardless of industry.

In addition to industry classification and VC backing, we examine the influence of firm size, post-IPO stock price performance, stock price volatility, the percentage of shares subject to lockup, the length of the lock-up period, secondary (or follow-on) offerings, underwriter reputation, trading volume, and other variables. We consistently find little or no reaction for the non-VC sample. For the VC sample, however, post-IPO price performance, abnormal trading volume, pre-expiration stock price volatility, and underwriter reputation are the most significant effects. The largest losses in value occur for firms with: (a) larger post-IPO stock price increases, (b) greater abnormal trading volume in the period surrounding lockup expiration, (c) greater pre-expiration stock price volatility, and (d) higher quality underwriters.

II. Background

A lockup provision is a contractual arrangement between insiders of a firm undergoing an IPO and the underwriter, in which insiders agree not to sell shares for a specified period, usually 180 days after the offer. Lockups are not required by law, but essentially all IPOs feature them. Insiders often own a large portion of the shares of a newly public firm. By restricting sales, the lockup agreement insures that insiders will maintain a significant economic interest in the firm after the IPO, thereby aligning the interests of old and new shareholders. Lockup agreements also limit the supply of shares available for trading, which may help support the issue price in the post-IPO period. Either way, the lockup agreement should increase the marketability of the IPO, thereby increasing its likelihood of success.

Lockups are not binding in that shares may be sold before expiration if consent is given by the underwriter. In addition, lockup expiration does not necessarily eliminate restrictions on insider sales. Insiders are still subject to Rule 144 and Rule 701, which place additional restrictions on insider trading (Rule 144 and Rule 701 restrictions are described in the Form S-1 excerpt in the Appendix). Furthermore, newly public companies, particularly in recent years, are often engaged in merger and acquisitions and/or other material nonpublic activity, further limiting insider selling possibilities. Thus, even though a lockup expires, it may be several years before an insider is legally allowed to sell shares.

To better illustrate some of the institutional features of lockup agreements, we briefly consider the case of Healtheon Corporation (now known as Healtheon/WebMD), which went public February 11, 1999, at $8 per share. The stock closed at $31.375, for a one-day gain of 292 percent. Information on lockup provisions is public knowledge and can be found in, for example, a company's Form S-1 under the heading, "Shares Eligible for Future Sale." The relevant portion of Healtheon's Form S-1 is reproduced in the Appendix. As shown, after the IPO, Healtheon would have a total of 67.196 million shares outstanding (assuming the underwriters did not exercise overallotment options and no other outstanding options or warrants were exercised); however, after the offering, only 5.658 million shares, including 5 million shares sold in the IPO, would be freely tradable.

Under the terms of Healtheon's underwriting agreement, 52.254 million shares were subject to a 180-day lockup. Of these shares, 41.817 million would still be subject to volume and other limitations of Rule 144 after this date, implying that 52.254 - 41.817 = 10.437 million shares would become tradable without restrictions of any type. The final 9.283 million shares would become available to sell at various dates after lockup expiration. Under the terms of the lockup agreement, Healtheon insiders are not only barred from selling the stock during the lockup period, but they are also barred from making any precommitment to sell and from entering into any transaction that transfers "the economic consequences of ownership."

Based on the information in Healtheon's S-1, the number of shares subject to the lockup exceeded the total shares sold to the public by a factor of about ten. Subject to Rule 144, an enormous increase in supply after the lockup period was possible. Healtheon's S-1 also reports that the underwriter, Morgan Stanley, has the right to release some of the locked up shares.

Healtheon's 180-day lockup period expired on August 10, 1999, at which time the stock was selling for $31.75, down from its intraday peak of $126 on May 21, 1999. On the lockup expiration day, Healtheon's share price dropped by more than 20 percent. Numerous sources reported on the decline, generally blaming the lockup expiration. For example, Reuters New Service (August 10, 1999) stated, "Shares of Healtheon Corp. fell as the online health care company's inside shareholders got their first chance to take profits since the initial public offering in February."

III. Data and Preliminary Analysis

Our primary data source for this study is the Securities Data Company (SDC) Global New Securities database. We obtain daily stock return data from the Center for Research in Security Prices (CRSP) file. The initial sample consists of all U.S. IPOs from January 1, 1988, through December 31, 1997. The starting date corresponds to the first entry in the SDC database containing a lockup provision, Jean Philippe Fragrances Inc., which completed its IPO January 15, 1988. Before this, SDC apparently did not capture data on lockup provisions. The ending date is based on our need for post-IPO stock market data from CRSP.

Sample Selection

According to SDC, 5,324 U.S. publicly traded initial common stock offerings occurred during our sample period. We first eliminated 1,990 offerings, primarily consisting of closed-end funds, depository shares, real estate investment trusts (REITs), reverse leveraged buyouts, spinoffs, and unit issues, leaving a potential sample of 3,334 firms. Of this group, 413 firms had no lockup agreement (according to SDC) and 115 firms had lockup periods that were too short for study. Another 6 firms had missing lockup periods, and 2 firms had reported secondary offerings before the IPO. Finally, as is common in IPO research, we eliminated 105 firms with offer prices less than $5. The final sample contains 2,693 firms.

Sample Characteristics

Figure I graphs the number of IPOs with lockup provisions from 1988 through 1997 as reported in the SDC database. In each year, IPOs are grouped based on whether the lockup period is less than 180 days, equal to 180 days, or greater than 180 days.

As Figure I shows, there is a trend toward standardizing lockup lengths at 180 days. Early in our sample, IPOs have lockups in roughly equal numbers in the three groups, but by the end of the sample, almost no IPOs have lockups shorter than 180 days, and 80 percent to 90 percent have lockups exactly equal to 180 days.

The trend toward standardized 180-day lockups has important implications for research such as that by Bray and Gompers (1999), who attempt to model the determinants of the length of the lockup period. They find, for example, that firms with greater informational asymmetries use longer lockups. However, because it appears that the variation in lockup lengths is sharply diminishing (and perhaps disappearing) through time, there is relatively little cross-sectional variation for such models to explain, at least in recent years. Why lockups lengths have become so standard is unclear, although Field and Hanka (2001) suggest it may be related to standardization of underwriting spreads as discussed in Chen and Ritter (2000).

The percentage of firms with lockup agreements is another concern. About 12 percent (413 of 3,334) of the firms we initially considered had no lockup agreement according to the SDC database. However, we examined S-1 filings for several dozen of these firms and found that, without exception, there actually was a lockup agreement. Thus, the incidence of lockup agreements is understated by SDC, perhaps significantly so. This is an important issue because some researchers explore factors that determine whether a lockup agreement is present in an IPO. If data errors are the primary reason for the apparent absence of a lockup, the results of such analyses may be misleading.

Table 1 provides greater detail on lockup agreements. Panel A examines the length of the lockup period for our sample of 2,693 firms. The average lockup period is 224 days; however, 75 percent (2,032 of 2,693) of the firms have lockup periods of exactly 180 days. Lockup periods of greater than one year are observed in 211 cases. A few firms have very long lockup periods, reaching a maximum of 1,095 days for 5 firms.

Panel B of Table 1 reports the percentage of shares subject to lockup, measured as the number of locked shares divided by shares outstanding after the IPO. The SDC database contains this information for 1,614 of the 2,693 firms in our sample. A relatively small number of firms are excluded from Panel B because this ratio exceeds 1.0, presumably because of data entry errors. As shown, among the firms in this subsample, the median is 63.30 percent. The 25th percentile is 51.60, implying that, for more than 75 percent of these firms, the number of locked shares exceeds the number of unlocked shares.

IV. Stock Price Reactions to IPO Lockup Expirations

Methods and Hypotheses

We employ standard event study methods using daily CRSP returns data. We use the CRSP value-weighted index in our market model estimations and examine abnormal returns using the standardized residual approach as in Patell (1976), Linn and McConnell (1983), and Schipper and Smith (1983). The lockup expiration day is day 0. We rely primarily on standard parametric z-statistics; however, we also calculate a nonparametric generalized binomial proportions test.

Our estimation period is seventy trading days, ending ten days before the event date. This estimation period represents a trade-off. A longer period is desirable for more accurate estimates, but, if only pre-event data are used, the length of the estimation period is limited by the relatively short lockup periods. In addition, we wished to exclude data from the post-IPO underwriter stabilization period. Of the 2,693 firms identified from the SDC database, we were not able to use 164 in our event study analysis, primarily because of differences in CUSIP numbers or missing returns data. Thus, our sample for the remainder of the article consists of the 2,529 firms with data in both sources.

The lockup expiration date is public knowledge; therefore, our null hypothesis is that there will be, on average, a zero abnormal return in all cases. Presumably, conditional on the information available to them, market participants form rational expectations regarding insider sales, and prices reflect those expectations. Furthermore, on the lockup expiration day, specific information about insider activities will not generally be available, so market participants are unable to directly observe insider activity.

Overall Sample Results

Table 2 provides the results of the overall event study analysis for the 2,529 firms in our sample for thirty days beginning on day -6. The average abnormal return (AR) on day 0 is -.74 percent, which is significant at the .0001 level. The cumulative abnormal return (CAR) for the day -2 to day +2 window is -1.61 percent, which is similarly significant. The median CAR for this period is -1.47 percent, and negative CARs outnumber positive CARs by about 1.5 to 1. The conclusions from the parametric results are supported by the proportions tests. Similar findings are reported by Brav and Gompers (1999), Field and Hanka (2001), and Ofek and Richardson (2000). Our conclusions are not sensitive to the choice of index or the event study methods used.

The results in Table 2 indicate lockup expirations are associated with negative abnormal returns. It is conceivable the observed negative abnormal returns are transient, because of price pressure or bid-ask bounce, particularly given that our sample contains many smaller, Nasdaq-listed firms. However, Figure II plots the CARs over the entire thirty-day period in Table 2. As illustrated, the prices show no tendency to rebound shortly after lockup expiration. Instead, there appears to be a permanent decline of about 2 percent, virtually all of which occurs in the days surrounding the expiration of the lockup.(1)

Partitioned Sample Analyses

The results in the previous subsection establish that IPO lockup expirations are, on average, associated with abnormal price declines. Although the effect is highly significant statistically, a price decline of 1 percent or so is much smaller than the typical bid-ask spread in our sample and probably too small to be profitably exploited (see Ofek and Richardson (2000) for a more detailed discussion of this issue). However, larger declines may exist for certain subgroups. Thus, our goal in this subsection is to determine whether the observed negative abnormal returns are concentrated in firms with certain characteristics. In subsequent subsections, we partition our sample based on various attributes and repeat the event study analysis.

Length of the Lockup Period. We first examine whether the length of the lockup period is a significant influence. We divide our sample into four groups based on the number of days in the lockup period using the divisions reported in Table 1. The results are in Table 3. As shown, firms with lockups of 180 days or less have a significant, negative CAR over the (-2,+2) window, and firms with lockups greater than 180 days generally have insignificant ARs, both on the event day and over a five-day window. Thus, the significant ARs are concentrated in firms with shorter lockup periods.

Industry Analysis. To explore the possibility of industry effects, we divide our sample based on one-digit Standard Industrial Classification (SIC) codes. Table 4 provides the results. As shown, our sample is not evenly spread across the groups. SIC codes 3 and 7 have the heaviest concentrations with 696 and 564 firms, respectively. SIC code 1 has only 70 firms. The remaining 5 are roughly even.

Examining the abnormal returns in Table 4, firms with SIC code 2 have the largest (in absolute value) AR on day 0 (-1.49 percent). The two largest (in absolute value) five-day CARs, -2.29 and -2.39 percent, occur for SIC codes 2 and 3, respectively. (2) The day 0 AR is negative in every case, but the event-day ARs and five-day CARs are generally not significant (at the 5 percent level) for industries 1, 4, and 6. Overall, based on the five-day CARs, SIC codes 2 and 3 appear to suffer the most severe decline, and SIC codes 1, 4, and 6 are essentially unaffected.

In the SDC database, certain firms are classified as high tech based on four-digit SIC codes, but the reporting appears to be inconsistent. Rather than rely on this variable, we obtained the relevant underlying codes directly from SDC and applied them to the firms in our sample. Although a relatively large number of codes are used, most fall under the two-digit classifications of 28 (biotechnology and drugs), 35 (computer and related), 36 (electronics and communication), 38 (medical equipment), and 73 (software). Using this classification, we compare the results for high-tech and non-high-tech firms in Table 5.

As shown, 1,048 firms in our sample, or about 40 percent, fall in the high-tech category. Examining the abnormal returns, we find that lockup expiration has a greater effect on these firms. The day 0 AR for high-tech firms is -1.2 percent, compared with -.42 percent for non-high-tech firms. Additionally, the five-day CAR for high-tech firms is more than triple that of non-high-tech companies (-2.75 percent versus -.80 percent). The most extreme losses appear to occur for companies classified as high-tech with a primary SIC code of 2. These firms lose more than 4 percent of their value in the period (-2,+2).

Venture Capital. Many IPOs feature VC backing, and several studies, including Barry et al. (1990), Bray and Gompers (1999), Hamao, Packer, and Ritter (2000), and Megginson and Weiss (1991) suggest possible differences between VC-backed and non-VC-backed firms. We therefore partition our sample into VC and non-VC firms in Table 6. A small number of firms have no classification in the SDC database and are omitted from this analysis.

As reported in Table 6, 1,137 firms, amounting to 45 percent of the sample, have VC backing. For these firms, the day 0 AR and five-day CAR are -1.25 percent and -2.81 percent, respectively. Although the ARs for the remaining 1,372 non-VC firms are negative and significant, they are about four times smaller. Thus, VC firms suffer much larger declines in value. Bray and Gompers (1999) and Field and Hanka (2001) also find more significant declines for VC-backed firms.

One possible explanation for the results in Table 6 is that VC and non-VC firms may have different characteristics. Table 7 presents basic summary statistics comparing VC and non-VC issues. The numbers shown are averages, with standard deviations in parentheses. The reported t-statistics test for differences in sample means.

We find the average offer sizes are similar at $37 million for VC firms and and $42 million for non-VC firms. The VC firms have larger post-IPO returns, but the difference is not dramatic, particularly for the 180-day returns. The VO firms have a slightly larger percentage of shares locked up. However, the difference in lockup period length is highly significant, and the shorter lockup lengths for VC firms helps explain our earlier finding that the negative abnormal returns are concentrated in firms with shorter lockup periods. Even so, when the analysis is restricted to VC and non-VC firms with lockups of exactly 180 days, the overall results are unaffected; the VC-backed firms have much larger losses in value. Overall, only two of the five means examined in Table 7 are statistically unequal (at the 1 percent level). Furthermore, with the possible exception of the length of the lockup period, the differences seem particularly large from an economic standpoint.

Comparing Tables 5 and 6, another possible explanation for the VC versus non-VC difference is that most of the firms in the high-tech group are VC. Barry et al. (1990) show that venture capitalists tend to focus or specialize in a subset of industries characterized as high tech; 63.8 percent of their sample is concentrated in computer equipment, electrical and electronic components, instrumentation, and business services. To explore whether the differences in Tables 5 and 6 are due to backing or industry (or both), we divide our sample into four groups based on VC versus non-VC and high tech versus non-high tech. The results are in Table 8.

As shown, VC firms suffer larger declines in value. The largest abnormal returns are associated with firms that are classified as both VC backed and high tech. This group has a day 0 AR of -1.59 percent and a (-2,+2) CAR of -3.33 percent. The next largest abnormal returns occur in the VC-backed, non-high-tech group, which has a day 0 AR of -.65 percent and a (- 2,+2) CAR of -1.91 percent. For the non-VC-backed firms, the high-tech group and non-high-tech group have similar day 0 ARs of -.34 percent and -.35 percent and (-2,+2) CARs of -1.47 percent and -.36 percent, respectively. Based on pairwise t-tests, both the AR and CAR for the VC, high-tech group are significantly larger than those in the other three groups.

We further examine these four groups by plotting their CARs over the (- 5,+23) window in Figure III. As illustrated, regardless of industry, VC firms show significant declines in value of about 4 percent whereas non-VC firms do not. In fact, the CAR over this period is essentially 0 for non-VC firms. Overall, the results in Table 8 and Figure III strongly suggest the presence or absence of VC backing is an important factor, whereas high-tech industry classification matters much less.

Our analysis thus far considers lockup dates across a ten-year period. In Figure IV, we present the (-2,+2) CARs for each year in our sample. Our goal is to determine whether the negative ARs persist as awareness of lockup expirations grows and to investigate whether the VC-backed firms consistently have greater losses.

The results in Figure IV show no evidence of a diminishing effect through time. For the overall sample, the CAR is significant and negative in seven of the ten years (Field and Hanka (2001) report a similar result). VC-backed firms have greater losses in all ten years. In fact, non-VC-backed firms have s significant and negative CAR in only two years, and the CAR is positive in three of the ten years. In contrast, VC-backed firms have a negative CAR every year, with highly significant values in eight of the ten years.

One explanation for this result is that VC firms have a comparative advantage when investing in complicated firms in their pre-IPO stages. Because the expertise required by VC firms is not easily obtained by market participants, venture capitalists act as gatekeepers for these investors. In addition, VC firms are not long-term, buy-and-hold investors. In this framework, the share price decrease we observe at lockup expiration is the result of portfolio rebalancing between VC investors and traditional equity investors, and the share price response is a liquidity event that occurs even though it is anticipated. (3)

Firm Size. Numerous studies identify differences among companies according to firm size. We partition our sample into deciles based on total market values and group our sample into VC-backed and non-VC-backed firms. Table 9 provides the results based on market values computed using the 180-day post-IPO price. (4)

For the overall sample, lockup expiration has the most pronounced effect for medium to larger firms (Brav and Gompers (1999) report a similar finding). As with our previous analyses, this result appears to be primarily due to the VC-backed firms; there seems to be no relation between market value and lockup expiration for the non-VC-backed firms.

That lockup expiration tends to be more important for larger firms is surprising. A priori, we might expect the reverse on the grounds that smaller firms tend to be riskier, have less analyst coverage, have less liquidity, and might be associated with greater asymmetric information. We explore a possible explanation in the next subsection.

Post-IPO Stock Price Performance. The market value measures in Table 9 are calculated based on stock prices 180 days after the offering. An explanation for the odd result may be that the larger firms include those with the greatest share price appreciation and that firm size per se is not the issue. To disentangle these two effects, we first calculate a firm's "IPO size" as the IPO offering price multiplied by shares outstanding after the IPO. Assuming no additional shares are sold after the IPO, the firm's market value at a later date is equal to its IPO size multiplied by 1 plus the percentage increase in the share price relative to the offering price. Table 10 presents the results obtained when we break firm size into these two components for the 180-day, post-IPO period. Because size does not seem to matter for non-VC firms, we consider only VC firms.

Panel A of Table 10 shows that when firm size is measured based on IPO size, there is no clear relation between size and lockup expiration. This conclusion is reinforced when we group firms based on total (book) assets in Panel C. However, when we form decile portfolios based on post-IPO stock price performance, in Panel B, we find better performers suffer a larger decline in share value at IPO lockup expiration. Based on the day 0 AR, decile 10, with an average post-IPO gain of 184.2 percent, is the hardest hit, losing an average of 2.15 percent on day 0 and more than 5 percent over the five-day surrounding period. The least affected firms are in decile 1. These firms, which have an average post-IPO loss of 53.3 percent, experience a statistically insignificant loss over the (-2,+2) period of -.63 percent. (5)

Volume Behavior. We examine the behavior of trading volume around lockup expiration. Figure V illustrates the average daily trading volume separately for VC and non-VC firms. Most noticeable is the spike in volume for the VC firms following lockup expiration, which peaks on day +1. For these firms, average volume at its peak is roughly double its pre-expiration level, and it remains approximately 30 percent higher. A much smaller, but similar, pattern exists for non-VO firms. Also, there is 50 percent to 100 percent greater average trading volume in the VC firms.

The results in Figure V (and in Tables 11 and 12) are consistent with the hypothesis that venture capitalists simply liquidate positions immediately after lockup expiration, which substantially increases the supply. Under this scenario, the abnormal price decline can be interpreted as evidence of downward-sloping demand curves for stock. However, as noted in Field and Hanka (2001), evidence on this point is difficult to obtain because neither distributions of shares to VC partners nor their subsequent sale must be disclosed (VC distributions are discussed in Gompers and Lerner (1998)).

Brav and Gompers (1999) conjecture that VC sales are the reason behind the volume spike and ARs because many VC firms must distribute shares once the lockup expires. Furthermore, most investors who receive distributed shares sell them automatically. In the absence of direct evidence on this point, Brav and Gompers and Field and Hanka (2001) perform similar analyses that examine whether the ARs around lockup expiration are related to the percentage of shares subject to lockup. Both studies find a significant and negative relation, and both interpret this result as supporting the hypothesis that the ARs are due, at least in part, to insider selling.

To investigate the behavior of trading volume, we calculate a measure of abnormal trading volume by dividing the average daily share volume during the (-2,+2) period by the average daily share volume over the seventy-day estimation period used in our event study. We then form decile portfolios based on abnormal volume. Table 11 shows the results.

Panel A presents results for the overall sample. The mean abnormal volume rises from .15 to 5.17, and the higher abnormal volume portfolios tend to experience the largest declines. Panels B and C repeat the analysis for VC and non-VC firms. As we consistently find, the effect is concentrated in VC firms. There is essentially no effect on non-VC firms even though decile average abnormal volumes range from .13 to 3.94. In fact, for decile 10, the five-day CAR is a significant 2.45 percent for non-VC firms.

In contrast, for the VC firms in Table 11, the largest losses are associated with decile 10. For this group, the average abnormal volume is 6.43, and the associated five-day CAR is -3.41 percent and highly significant. In contrast, the decile 1 abnormal volume is .17, and the ARs are statistically insignificant.

In Table 12, we analyze the joint influence of post-IPO price performance and abnormal volume by dividing our sample into four groups based on whether abnormal volume is greater or less than than 1.0 and whether post-IPO performance is positive or negative. The results of this analysis are reported separately for the VC firms (Panel A) and non-VC firms (Panel B).

Panel A shows the 361 VC firms experience positive post-IPO stock price performance and abnormal volume greater than 1.0. Over the (-2,+2) period, these firms lose an average of 3.75 percent of their value, which is the most significant loss among the four groups considered. In contrast, the 252 VC firms with negative post-IPO performance and abnormally low volume experience a statistically insignificant -.56 percent decline.

The remaining two VC groups are similar in that the five-day CAR is -3.28 percent for the low-performance, high-volume group and -3.21 percent for the high-performance, low-volume group. In contrast to the VC firms in Panel A, Panel B shows that neither performance nor volume consistently matters for the non-VC firms. In fact, for the high-performance, high-volume group, the day 0 AR and five-day CAR are positive, albeit insignificantly so.

Taken together, Tables 9 through 11 show that lockup expiration typically has a relatively small effect on non-VC firms. For VC firms, lockup expirations are associated with significant and negative ARs, especially for firms that experience post-IPO price increases. The losses are particularly pronounced for these firms when there is abnormally high volume at lockup expiration.

VI. Follow-On Offerings: A Brief Analysis

Many of the firms in our sample have a seasoned equity offering (SEO), also known as a secondary or follow-on offering, during the sample period. On some occasions, such offers are put together to facilitate insider sales. Presumably, the goal in such cases is to lessen the effect of large-scale insider selling by allowing the underwriter to market the issue and to alleviate concerns about seller motivation.

We examine our sample to determine whether SEOs are common around the time of lockup expirations and whether SEOs are more common for VC firms. A total of 890 firms, about one-third of the sample, filed for an SEO by the end of our sample period, and 220 of the 890 firms filed for an SEO before the lockup expiration date.

Comparing SEO filings for VC firms and non-VC firms, we find the two groups are similar in terms of the number of SEOs filed and the number filed near the lockup expiration date. However, there are almost 20 percent more non-VC firms than VC firms in our sample, so the VC firms file for SEOs more frequently on a percentage basis.

To determine whether SEO filings affect the results in previous sections, we repeat much of the analysis in this study after eliminating firms that file for SEOs any time in the (-100,+3) window. We find no notable differences in the results.

VII. Multivariate Analyses

We now turn to multiple regression analysis to consider the joint effects of issue and firm characteristics investigated in previous sections. These regressions facilitate comparison of our results with other studies, particularly Brav and Gompers (1999) and Field and Hanka (2001). To further enhance comparability, we add variables examined in other studies.

In our regressions, the five-day CAR is the dependent variable. The independent variables are:

VC = dummy variable equal to 1 if firm is VC backed, 0 otherwise;

SDCTECH = dummy variable equal to 1 if firm is high tech, 0 otherwise;

PER180 = 180-day, post-IPO stock price performance;

RATIO_V = abnormal volume as defined in section IV;

BIG3 = dummy variable equal to 1 if firm's underwriter is one of the "big three," (6) 0 otherwise;

SD = standard deviation of market model residuals from the 70-day event-study estimation period;

LNSIZE = natural logarithm of the total market capitalization based on the offering price and shares outstanding after the IPO;

LOCKTIME = length of the lockup period in days;

SEO = dummy variable equal to 1 if firm has an SEO within the sample period, 0 otherwise; and

PERLOCK = percentage of shares outstanding subject to lockup as defined in section III.

Examining the whole sample results in Panel A, the VC and high-tech dummies are consistently negative and significant, as are the 180-day stock price performance and, to a lesser extent, underwriter quality. Firm size is significant and positive (at the 5 percent level), indicating larger firms suffer smaller declines in value after controlling for performance. Abnormal volume, the length of the lockup, whether the firm does an SEO, and the residual standard deviation (with one exception) do not appear to be significant.

The results of this analysis, with p-values in parentheses, are presented in Table 13.

In the last regression in Panel A, the percentage of shares subject to lockup is included, which reduces the sample size by about 40 percent. Consistent with Brav and Gompers (1999) and Field and Hanka (2001), the coefficient is negative and significant (p-value = .021), indicating firms with a greater percentage of shares locked up suffer larger declines in value. Except for the residual standard deviation, the results for the other variables are similar to the larger sample regressions.

Because much of our previous analysis suggests that VC and non-VC firms may be subject to different influences, we examine the two groups separately in Panel B (VC firms) and Panel C (non-VC firms). The regressions are similar to those in Panel A, except the VC dummy is removed. Based on a standard F-test (a Chow test), we reject the equality of the fitted regressions in every case at conventional significance levels. We conclude there are significant differences in the VC and non-VC backed samples, and pooling may be inadvisable.

When the individual coefficient results for the two groups are compared, some statistically significant differences emerge. For example, abnormal volume is insignificant in the overall sample; however, it is significant and negative for VC firms and significant and positive for non-VC firms. The residual standard deviation is generally not significant in the overall regressions, but it is generally significant and negative for VC firms and significant and positive for non-VC firms.

In addition, 180-day performance and underwriter quality are significant only for VC firms, whereas size is significant only for non-VC firms. The tech dummy is more significant for non-VC firms, as well. The length of the lockup and the presence of an SEO do not appear to matter for either group. Finally, the percentage of shares that are locked up has a similar coefficient in both cases, but it is significant only for non-VC firms. Thus, we cannot conclude that VC firms with the greatest percentage of locked shares will suffer larger losses when the lockup expires.

VIII. Summary and Conclusions

Most IPOs feature lockup agreements. Such agreements bar insiders from selling the stock for a set period after the IPO, usually 180 days. The lockup expiration date thus represents the first opportunity for insiders to sell in the secondary market. Because the percentage of a firm's shares subject to lockup is often 100 percent or more, a large, sudden increase in supply is possible. Significant share price revisions may occur if market participants infer private information from perceived insider transactions.

To evaluate the effects of lockup expiration, we examine stock price behavior in the period surrounding the lockup expiration date for a sample of 2,529 firms from 1988 to 1997. We find that lockup expirations are, on average, associated with significant and negative abnormal returns. We further find that the negative abnormal returns in the period surrounding lockup expiration are mostly due to the 45 percent of the firms in our sample with VC backing. Such firms lose, on average, 3 percent to 4 percent percent of their value in this period, and high-tech firms with VC backing are particularly hard hit. Non-VC-backed firms lose relatively little value, regardless of industry.

In addition to industry and VC backing, we examine the influence of firm size, post-IPO stock price performance, underwriter reputation, stock price volatility, the percentage of shares subject to lockup, the length of the lock-up period, secondary (or follow-on) offerings, trading volume, and other variables. We find little or no reaction for the non-VC-backed sample. For the VC-backed sample, post-IPO price performance and trading volume are the most significant effects. The largest losses in value occur for firms with the largest stock price increases, firms that experience abnormally high trading volume in the period surrounding lockup expiration, and firms with greater stock price volatility during the pre-expiration period. Firms associated with high-quality underwriters also appear to sustain larger losses in the period surrounding lockup expiration.

The results in this study raise some public policy issues. Although the relevant date is, in principle, public knowledge, should firms be required to inform shareholders just before expiration? At a minimum, it would seem reasonable for firms to disclose, in advance, when a lockup is going to be released earlier than originally scheduled. Similarly, should insiders be required to disclose, in advance, their intentions once the lockup expires? Whether such disclosures would eliminate the losses suffered by shareholders around lockup expiration is an open question, but the possibility exists.

We thank Bill Megginson, Bob Miller, Dick Pettway, and Jay Ritter, along with seminar participants at Miami University, the University of Kansas, the University of Kentucky, and Wright State University, and an anonymous referee.

(1.) Field and Hanka (2001) and Ofek and Richardson (2000) examine bid and ask prices around lockup expiration. Both studies report essentially parallel shifts, which supports this conclusion. Also, Field and Hanka find no evidence that earnings announcements systematically occur near lockup expirations.

(2.) To avoid repetitious language, we omit the qualifier "in absolute value" when the meaning is clear in context.

(3.) We thank an anonymous reviewer for suggesting this explanation.

(4.) In Tables 9 through 11, we report results only for deciles 1, 4, 7, and 10 to save space. The full results are available on request.

(5.) The 180-day return includes the IPO initial return. We repeated this analysis using just the initial return and found a similar, but less distinct, pattern.

(6.) The "big three" underwriters are Goldman Sachs, Merrill Lynch, and Morgan Stanley. Over our sample period, these three are the dominant underwriters based on equity IPO market share. A detailed analysis of underwriter quality in this context is available from the authors on request.

References

Barry, C. B., C. J. Muscarella, J. W. Peavy III, and M. R. Vetsuypens, 1990, The role of venture capital in the creation of public companies: Evidence from the going-public process, Journal of Financial Economics 27, 447-71.

Brau, J., 1999, An empirical analysis of the IPO lockup period, Working paper, Brigham Young University.

Brav, A. and P. A. Gompers, 1999, Insider trading subsequent to initial public offerings: Evidence from expiration of lock-up provisions, Working paper, Duke University and Harvard University.

Chen, H. C. and J. R. Ritter, 2000, The seven percent solution, Journal of Finance 55, 1105-31.

Field, L. C. and G. Hanka, 2001, The expiration of IPO share lockups, Journal of Finance 56, 471-500.

Gompers, P. and J. Lerner, 1998, Venture capital distributions: Short-run and long-run reactions, Journal of Finance 53, 2161-83.

Hamao, Y., F. Packer, and J. R. Ritter, 2000, Institutional affiliation and the role of venture capital: Evidence from initial public offerings in Japan, Pacific Basin Finance Journal 8, 529-58.

Linn, S. C. and J. J. McConnell, 1983, An empirical investigation of the impact of "antitakeover" amendments on common stock prices, Journal of Financial Economics 11, 361-99.

Megginson, W. and K. Weiss, 1991, Venture capitalist certification in initial public offerings, Journal of Finance 46, 879-903.

Ofek, E. and M. Richardson, 2000, The IPO lock-up period: Implications for market efficiency and downward sloping demand curves, Working paper, New York University.

Patell, J. M., 1976, Corporate forecasts of earnings per share and stock price behavior: Empirical tests, Journal of Accounting Research 14, 246-76.

Schipper, K. and A. Smith, 1983, Effects of recontracting on shareholder wealth: The case of voluntary spin-offs, Journal of Financial Economics 12, 437-68.

[Graph omitted]

[Graph omitted]

[Graph omitted]

[Graph omitted]

[Graph omitted]

TABLE 1.

Sample Characteristics.

Panel A. Length of Lockup in Calendar Days (N = 2,693)

                         Lockup Period
           120 [less than or equal to] x   180
                       < 180

N                       164               2,032
Mean                    130                 180
Median                  120                 180
Mode                    120                 180
Std. dev.                14                   0

                             Lockup Period
           180 < x [less than or equal  > 365      All
                     to] 365

N                      286               211      2,693
Mean                   328               589        224
Median                 360               545        180
Mode                   365               730        180
Std. dev.               53               154        125
Panel B. Percentage of Shares Subject to Lockup (N = 1,614)

Percentile  Mean    25%    50%    75%    99%

            59.54  51.60  63.30  71.74  88.77

Note: This table provides descriptive statistics for the length of
lockup provisions and percentage of shares subject to lockup. Panel A
reports the lockup provision in days for selected ranges of the lockup
provision. Panel B reports summary statistics on the fraction of shares
subject to lockup provisions. This variable is calculated as the number
of shares locked divided by the number of shares outstanding after the
offer. The data are from the Securities Data Company Global New
Securities database from 1988 through 1997.
TABLE 2.

Event Study Results: Entire Sample.

Panel A. Abnormal Returns

     Average    Median                           Generalized
     Abnormal  Abnormal               Postive:      Sign
Day   Return    Return     z     N    Negative        z

-6     0.12%    -0.06%    0.69  2529  1237:1292      1.30
-5     0.05     -0.12     0.52  2529  1199:1330     -0.21
-4     0.05     -0.11     0.25  2529  1205:1324      0.03
-3    -0.39     -0.29    -4.16  2529  1127:1402     -3.08
-2    -0.11     -0.13    -1.11  2529  1196:1333     -0.33
-l    -0.32     -0.34    -4.60  2529  1091:1438     -4.51
 0    -0.74     -0.57    -9.22  2529  1045:1484     -6.34
 1    -0.34     -0.37    -4.99  2528  1122:1406     -3.26
 2    -0.10     -0.15    -0.52  2528  1185:1343     -0.75
 3     0.00     -0.16    -0.55  2528  1183:1345     -0.83
 4    -0.01     -0.09    -0.08  2528  1221:1307      0.68
 5    -0.15     -0.17    -2.26  2528  1171:1357     -1.31
 6    -0.05     -0.17    -1.46  2528  1167:1361     -1.47
 7     0.13     -0.11     1.11  2528  1204:1324      0.01
 8     0.06     -0.14     0.06  2528  1180:1348     -0.95
 9    -0.08     -0.15    -1.28  2528  1203:1325     -0.03
10    -0.07     -0.17    -1.60  2528  1175:1353     -1.15
11    -0.05     -0.14    -0.77  2528  1190:1338     -0.55
12    -0.03     -0.07    -0.51  2528  1224:1304      0.80
13    -0.05     -0.17     0.02  2526  1177:1349     -1.03
14     0.07     -0.05     0.12  2525  1230:1295      1.10
15    -0.02     -0.07    -0.03  2524  1219:1305      0.68
16    -0.18     -0.19    -2.21  2524  1163:1361     -1.55
17     0.12     -0.07     0.89  2524  1223:1301      0.84
18     0.02     -0.13     0.87  2523  1203:1320      0.06
19    -0.17     -0.13    -1.86  2522  1196:1326     -0.20
20     0.06     -0.12    -0.08  2522  1203:1319      0.08
21     0.17     -0.10     1.24  2521  1200:1321     -0.02
22    -0.10     -0.07    -0.52  2520  1222:1298      0.88
23     0.10     -0.07     0.84  2520  1207:1313      0.28
Panel B. Cumulative Abnormal Returns

                   CAR

          Equally   Precision  Median              Positive:
Days      Weighted  Weighted    CAR        z       Negative

(-1,+1)    -1.39%    -1.34%    -1.35%    -10.86    1002:1527
(-2,+2)    -1.61     -1.45     -1.47      -9.14    1031:1497


          Generalized
             Sign
Days           z

(-1,+1)      -8.06
(-2,+2)      -6.90

Note: This table reports the event study results for the entire sample
around lockup expiration. The standardized residual method and
value-weighted index are used to compute and evaluate abnormal returns.
Day 0 is the lockup expiration day. The windows (-1,+1) and (-2,+2) are
reported. The generalized sign z tests the null hypothesis that the
percentage of positive returns is the same as in the estimation period.
The data are from the Securities Data Company Global New Securities
database from 1988 through 1997.
TABLE 3.

Event Study Based on Length of Lockup Period.

Lockup Period                          N    Day 0 AR    z

120 [less than or equal to] x < 180    158    -.24%    -1.09
180                                  1,947    -.93    -10.21
180 < x [less than or equal to] 365    268    -.29      -.53
x > 365                                156     .38       .75

Lockup Period                        (-2,+2) CAR    z

120 [less than or equal to] x < 180     -2.55%    -4.39
180                                     -1.70     -8.36
180 < x [less than or equal to] 365     -1.01     -1.60
x > 365                                  -.57      -.75

Note: Day 0 AR is the event-day abnormal return followed by the
corresponding z-statistic. (-2,+2) CAR is the cumulative abnormal return
in the (-2,+2) window followed by the corresponding z-statistic. The
standardized residual method and value-weighted index are used to
compute and evaluate abnormal returns. The data are from the Securities
Data Company Global New Securities database from 1988 through 1997.
TABLE 4.

Event Study Results for Industry Classifications.

SIC One-Digit Code          N   Day 0 AR    z    (-2,+2) CAR    z

1 Mining and construction   70    -.58%   -1.76     -.24%      -.29
2 Light manufacturing      287   -1.49    -5.15    -2.29      -3.92
3 Heavy manufacturing      696   -1.06    -6.69    -2.39      -6.34
4 Regulated industries     194    -.40    -1.11      .13       -.32
5 Wholesale and retail     300    -.61    -2.41    -1.74      -3.41
6 Financials               182    -.27    -1.52     0.20       -.73
7 Service                  564    -.37    -2.88    -1.58      -4.38
8 Health service           232    -.65    -2.65    -1.59      -3.01

Note: This table provides the results of an industry analysis based on
one-digit SIC codes around lockup expiration. The standardized residual
method and value-weighted index are used to compute and evaluate
abnormal returns. Day 0 AR is the event-day abnormal return followed by
the corresponding z-statistic. (-2,+2) CAR is the cumulative abnormal
return in the (-2,+2) window followed by the corresponding z-statistic.
We exclude industries with one-digit SIC codes of 0 and 9 because the
sample size was fewer than fifty observations. The data are from the
Securities Data Company Global New Securities database from 1988 through
1997.
TABLE 5.

Event Study Results for High-Tech Firms.

Type of Firm            N    Day 0 AR    z    (-2,+2) CAR    z

Non-high tech         1,481   -.42%    -4.09      -.80%    -4.17
High tech             1,048  -1.20     -9.46     -2.75     -9.24
High tech SIC code 2    135  -2.55     -6.24     -4.13     -4.73

Note: This table shows the difference between high-tech and
non-high-tech firms around lockup expiration. High-tech firms are
generated using a four-digit SIC code provided by Securities Data
Company. The standardized residual method and value-weighted index are
used to compute and evaluate abnormal returns. Day 0 AR is the event-day
abnormal return followed by the corresponding z-statistic. (-2,+2) CAR
is the cumulative abnormal return around the window (-2,+2) followed by
the corresponding z-statistic. The data are from the Securities Data
Company Global New Securities database from 1988 through 1997.
TABLE 6.

Venture-Capital-Backed Versus Non-Venture-Capital-Backed Firms.

Type of Firm                  N    Day 0 AR    z    (-2,+2) CAR    z

Venture capital backed      1,137   -1.25%   -9.51    -2.81%     -10.31
Non-venture capital backed  1,372    -.35    -3.95     -.62       -2.96

Note: The standardized residual method and value-weighted index are used
to compute and evaluate abnormal returns. Day 0 AR is the event-day
abnormal return followed by the corresponding z-statistic. (-2,+2) CAR
is the cumulative abnormal return around the window (-2,+2) followed by
the corresponding z-statistic. The data are from the Securities Data
Company Global New Securities database from 1988 through 1997.
TABLE 7.

Descriptive Statistics for Venture Capital Firms Versus Non-Venture-
Capital Firms.

                                Venture         Non-Venture
Variable                   N    Capital    N      Capital    t-stat

Offer amount (mil $)     1,156   37.19   1,506     42.00      -2.27
                                (39.59)           (63.05)
90-day performance (%)   1,127   27.38   1,434     20.20       3.86
                                (52.16)           (42.00)
180-day performance (%)  1,137   29.40   1,450     23.97       1.97
                                (72.12)           (67.09)
Shares locked (%)          784   60.47     821     58.66       2.06
                                (16.70)           (18.40)
Average number of        1,160  191.62   1,511    249.30     -12.18
days in lockup period           (65.07)          (150.83)

Note: Offer amount is the offer price times the number of shares offered
to the market. Performance is calculated as the 90-day and 180-day stock
price minus the offer price divided by the offer price. Standard
deviations are in parentheses. The data are from the Securities Data
Company Global New Securities database from 1988 through 1997.
TABLE 8.

Event Study Comparison of Venture-Capital-Backed (Non-Venture-Capital-
Backed) and High-Tech (Non-High-Tech) Firms.

               Venture Capital  Non-Venture Capital

High tech

 Day 0 AR          -1.59%              -.34%
 (z)              (-9.52)            (-2.78)
 (-2,+2) CAR       -3.33%             -1.47%
 (z)              (-9.08)            (-3.01)
 N                724                322

Non-high tech

 Day 0 AR           -.65%              -.35%
 (z)              (-3.18)            (-2.98)
 (-2,+2) CAR       -1.91%              -.36%
 (z)              (-5.08)            (-1.71)
 N                413              1,050

Note: High-tech firms are generated using a four-digit SIC code provided
by Securities Data Company. The standardized residual method and
value-weighted index are used to compute and evaluate abnormal returns.
Day 0 AR is the event-day abnormal return followed by the corresponding
z-statistic. (-2,+2) CAR is the cumulative abnormal return around the
window (-2,+2) followed by the corresponding z-statistic. The data are
from the Securities Data Company Global New Securities database from
1988 through 1997.
TABLE 9.

Event Study Results Based on Market Value for Whole, Venture Capital,
and Non-Venture Capital Samples.

Panel A. Entire Sample

Decile   Day 0 AR    z    (-2,+2) CAR    z     Mean

 1         -.37%    -.22      -.97%     -.42  $13.2
 4         -.73    -2.23     -2.29     -4.36   61.4
 7        -1.41    -6.01     -2.37     -4.49  140.0
10         -.59    -2.04     -1.54     -1.84  885.4

Panel B. Venture Capital

Decile   Day 0 AR    z    (-2,+2) CAR    z     Mean

 1         -.91%    -.98     -2.66%    -2.32  $21.7
 4        -1.85    -4.23     -1.90     -2.40   74.5
 7        -1.85    -4.74     -2.82     -3.41  148.1
10        -2.13    -4.92     -4.91     -5.19  756.6

Panel C. Non-Venture Capital

Decile   Day 0 AR    z    (-2,+2) CAR    z     Mean

 1         -.05%     .27      -.12%      .18  $10.5
 4         -.05     -.81     -1.10     -1.58   50.6
 7         -.83    -2.92     -1.63     -2.32  134.5
10          .13      .26       .47      1.32  989.7

Note: Market value measures are calculated by multiplying the number of
shares outstanding after the offer by the stock price for each firm.
Deciles are increasing with market value (i.e., smallest firms in decile
1). The standardized residual method and value-weighted index are used
to compute and evaluate abnormal returns. Day 0 AR is the event-day
abnormal return followed by the corresponding z-statistic. (-2,+2) CAR
is the cumulative abnormal return in the (-2,+2) window followed by the
corresponding z-statistic. Means are expressed in millions of dollars.
Market value measures are based on the 180-day post-IPO stock price. The
data are from the Securities Data Company Global New Securities database
from 1988 through 1997.
TABLE 10.

Event Study Results for Venture-Capital-Backed Decile Portfolios Based
on IPO Size, 180-Day Performance, and Total Assets.

Panel A. IPO Size

Decile  Day 0 AR    z    (-2,+2) CAR    z     Mean

 1      -1.52%    -2.03    -4.09%     -3.75  $25.2
 4      -1.82     -4.35    -3.27      -3.84   68.9
 7       -.83     -1.50    -2.96      -3.12  122.2
10      -1.73     -4.32    -4.04      -4.59  501.2

Panel B. 180-Day Performance

Decile  Day 0 AR    z    (-2,+2) CAR    z     Mean

 1       -.57%    -1.25     -.63%      -.51  -53.3%
 4      -1.26     -3.05    -2.56      -2.91   -4.5
 7      -1.18     -2.94    -2.69      -3.27   40.0
10       -.36     -1.16    -1.71      -2.80  493.4

Panel C. Total Assets

Decile  Day 0 AR    z    (-2,+2) CAR    z     Mean

 1      -1.15%    -1.59    -3.82%     -3.52  $11.8
 4       -.74     -1.57    -1.28      -1.84   30.5
 7      -1.23     -3.37    -2.62      -2.95   50.0
10      -2.15     -4.79    -5.03      -4.91  184.2

Note: IPO size is calculated by multiplying the number of shares
outstanding after the IPO by the offer price. The 180-day performance is
measured as the 180-day return relative to the offer price. Total assets
and IPO size are in millions of dollars. The standardized residual
method and value-weighted index are used to compute and evaluate
abnormal returns. Day 0 AR is the event-day abnormal return followed by
the corresponding z-statistic. (-2,+2) CAR is the cumulative abnormal
return in the (-2,+2) window followed by the corresponding z-statistic.
The data are from the Securities Data Company Global New Securities
database from 1988 through 1997.
TABLE 11.

Events Study Results Based on Relative Volume for whole, Venture
Capital, and Non-Venture Capital Samples.

Panel A. Entire Sample

Decile  Day 0 AR    z    (-2,+2) CAR    z    Mean

 1        -.26%   -1.25      -.54%    -1.32   .15
 4        -.02     -.52     -1.99     -2.99   .57
 7        -.37    -1.93     -1.82     -3.01  1.18
10       -1.01    -4.09      -.82     -2.09  5.17

Panel B. Venture Capital

Decile  Day 0 AR    z    (-2,+2) CAR    z    Mean

 1        -.56%   -1.08     -1.35%    -1.61   .17
 4        -.53    -1.05     -2.42     -2.78   .67
 7       -1.46    -3.88     -2.53     -3.37  1.46
10       -1.37    -3.20     -3.41     -4.20  6.43

Panel C. Non-Venture Capital

Decile  Day 0 AR    z    (-2,+2) CAR    z    Mean

 1         .02%    -.53       .11%     -.13   .13
 4        -.14     -.49     -1.29     -1.57   .49
 7        -.79    -2.39      -.96     -1.25  1.02
10        -.41    -1.59      2.45      2.94  3.94

Note: Relative volume is defined as the ratio of the average volume
during the event period (-2,+2) divided by the average volume during the
70-day estimation period. Deciles are increasing with relative trading
volume (i.e., the smallest relative volume is in decile 1). The
standardized residual method and value-weighted index are used to
compute and evaluate abnormal returns. Day 0 AR is the event-day
abnormal return followed by the corresponding z-statistic. (-2,+2) CAR
is the cumulative abnormal return in the (-2,+2) window followed by the
corresponding z-statistic. The data are from the Securities Data Company
Global New Securities database from 1988 through 1997.
TABLE 12.

Event Study Results of the Relation Between 180-Day, Post-IPO
Performance, and Relative Volume for Venture-Capital and
Non-Venture-Capital Firms.

Panel A. Venture Capital

                                   Postive 180-Day  Negative 180-Day
                                     Performance      Performance

Relative volume ratio greater
than 1
 Day 0 AR                              -1.82%            -1.98%
 (z)                                  (-7.76)           (-6.69)
 (-2,+2) CAR                           -3.75%            -3.21%
 (z)                                  (-7.47)           (-5.23)
 N                                    361               193
Relative volume ratio less than 1
 Day 0 AR                               -.93%             -.27%
 (z)                                  (-3.89)            (-.60)
 (-2,+2) CAR                           -3.28%             -.56%
 (z)                                  (-6.61)            (-.80)
 N                                    331               252

Panel B. Non-Venture Capital

                                   Postive 180-Day  Negative 180-Day
                                     Performance      Performance

Relative volume ratio greater
than 1
 Day 0 AR                                .31%            -1.01%
 (z)                                   (-.27)           (-3.35)
 (-2,+2) CAR                             .55%             1.38%
 (z)                                   (1.89)             (.71)
 N                                    324               171
Relative volume ratio less than 1
 Day 0 AR                               -.30%             -.71%
 (z)                                  (-2.42)           (-2.79)
 (-2,+2) CAR                           -1.77%             -.93%
 (z)                                  (-4.60)           (-1.54)
 N                                    539               338

Note: Relative volume is defined as the ratio of the average volume
during the event period (-2,+2) divided by the average volume during the
70-day estimation period. Performance is measured as the 180-day return
based on the offer price. The standardized residual method and
value-weighted index are used to compute and evaluate abnormal returns.
Day 0 is the event-day abnormal return followed by the corresponding
z-statistic. (-2,+2) CAR is the cumulative abnormal return around the
window (-2,+2) followed by the corresponding z-statistic. The data are
form the Securities Data Company Global New Securities database from
1988 through 1997.
TABLE 13.

Cross-Sectional Regressions for Five-Day Cumulative Abnormal
Returns.

Panel A. Full Sample Results

Regression   N    Intercept    VC     SDCTECH  PER180  RATIO_V  BIG3

(1)         2458    -.05     -1.57    -1.23    -.01     .09     -1.04
                    (.879)    (.000)   (.004)  (.000)  (.400)    (.087)
(2)         2458     .01     -1.57    -1.21    -.01     .09     -1.05
                    (.990)    (.000)   (.006)  (.000)  (.405)    (.087)
(3)         2456   -9.47     -1.62    -1.27    -.01     .09     -1.67
                    (.027)    (.000)   (.004)  (.000)  (.387)    (.012)
(4)         2456  -10.31     -1.58    -1.26    -.01     .09     -1.70
                    (.029)    (.000)   (.005)  (.000)  (.394)    (.011)
(5)         2456  -10.35     -1.58    -1.26    -.01     .09     -1.70
                    (.029)    (.000)   (.005)  (.000)  (.393)    (.011)
(6)         1494  -17.15     -1.24    -1.66    -.01     .00     -1.77
                    (.008)    (.028)   (.004)  (.035)  (.986)    (.050)

Regression    SD    LNSIZE  LOCKTIME   SEO    PERLOCK  Adj. [R.sup.2]

(1)                                                         .023

(2)         -.13                                            .022
            (.898)
(3)          .50     .51                                    .024
            (.635)  (.026)
(4)          .36     .55      .00                           .024
            (.741)  (.025)   (.665)
(5)          .36     .55      .00     -.06                  .023
            (.740)  (.025)   (.661)   (.931)
(6)         3.41     .96      .00     -.20    -.03          .025
            (.019)  (.005)   (.570)   (.829)  (.021)
Panel B. Venture Capital Results

Regression   N    Intercept  SDCTECH  PER180  RATIO_V     BIG3       SD

(1)         1121     -.65    -1.30    -.02     -.33    -1.66
                     (.233)   (.034)  (.000)   (.015)   (.047)
(2)         1121     1.77     -.77    -.02     -.37    -1.93    -5.87
                     (.064)   (.226)  (.000)   (.006)   (.021)   (.002)
(3)         1121    -3.29     -.78    -.02     -.37    -2.20    -5.71
                     (.675)   (.220)  (.000)   (.007)   (.018)   (.003)
(4)         1121     -.18     -.82    -.02     -.36    -2.16    -5.37
                     (.983)   (.195)  (.000)   (.007)   (.021)   (.006)
(5)         1121      .79     -.82    -.02     -.36    -2.11    -5.42
                     (.925)   (.196)  (.000)   (.008)   (.024)   (.006)
(6)         754     -6.22    -1.19    -.01     -.35    -1.94      .75
                     (.549)   (.133)  (.011)   (.018)   (.093)   (.773)

Regression  LNSIZE  LOCKTIME      SEO  PERLOCK   Adj. [R.sup.2

(1)                                                 .028

(2)                                                 .036

(3)          .27                                    .035
            (.515)
(4)          .16      -.01                          .036
            (.713)    (.232)
(5)          .11      -.01      .85                 .035
            (.804)    (.207)   (.424)
(6)          .43      -.01      .59     -.03        .020
            (.437)    (.369)   (.646)   (.242)

Panel C. Non-Venture Capital Results

Regression   N    Intercept  SDCTECH  PER180  RATIO_V     BIG3       SD

(1)         1337    -1.35    -1.14    -.01     1.05     -.17
                     (.000)   (.059)  (.166)   (.000)   (.846)
(2)         1337    -2.43    -1.44     .00     1.06      .12     2.54
                     (.000)   (.020)  (.363)   (.000)   (.896)   (.029)
(3)         1335   -13.74    -1.55     .00     1.06     -.83     3.42
                     (.006)   (.012)  (.314)   (.000)   (.397)   (.005)
(4)         1335   -15.27    -1.53     .00     1.05     -.90     3.16
                     (.006)   (.014)  (.308)   (.000)   (.362)   (.015)
(5)         1335   -15.69    -1.57     .00     1.08     -.85     3.18
                     (.005)   (.011)  (.616)   (.000)   (.385)   (.014)
(6)         740    -22.67    -2.21     .00     1.47    -1.02     4.35
                     (.005)   (.008)  (.914)   (.00)    (.486)   (.013)

Regression  LNSIZE  LOCKTIME      SEO  PERLOCK   Adj. [R.sup.2

(1)                                                 .023

(2)                                                 .026

(3)          .61                                    .029
            (.024)
(4)          .68       .00                          .029
            (.020)    (.530)
(5)          .70       .00    -1.44                 .030
            (.016)    (.463)   (.161)
(6)         1.16       .00    -1.94     -.04        .050
            (.008)    (.235)   (.152)   (.028)

Note: This table provides the results of ordinary least squares
regressions, with p-values in parentheses, of (-2,+2) day cumulative
abnormal returns against a dummy variable equal to 1 if the issue is
venture capital backed, 0 otherwise (VC); a dummy variable equal to 1 if
the issue is high tech, 0 otherwise (SDCTECH); 180-day post-IPO
performance (PER180); abnormal volume (RATIO_V); a dummy variable equal
to 1 if the issue was underwritten by a big three underwriter, 0
otherwise (BIG3); the residual standard deviation over the 70-day
estimation period (SD); the log of the amount offered (LNSIZE); the
number of days a firm is subject to lockup provisions (LOCKTIME); a
dummy variable equal to 1 if the firm partcipated in a follow-on
offering, 0 otherwise (SEO); and the fraction of shares locked to total
shares outstanding (PERLOCK). High-tech firms are generated using a
four-digit SIC code provided by Securities Data Company. The 180-day
performance is calculated as the stock price 180 days after the offer
minus the offer price divided by the offer price. Abnormal volume is
calculated as the average (-2,+2) window volume dividied by the 70-day
average volume during the estimation period. Morgan Stanley, Goldman
Sachs, and Merrill Lynch represent the big three underwriters. The
fraction of shares subject to lockup provisions is calculated as the
number of shares locked divided by the number of shares outstanding. The
data are from the Securities Data Company Global New Securities database
from 1988 through 1997.

Appendix

Healtheon Corporation's Lockup Agreement

The following is excerpted from Healtheon Corporation's Form S-1, filed February 10, 1999.

SHARES ELIGIBLE FOR FUTURE SALE

Prior to this offering, there has been no public market for the common stock of Healtheon. Future sales of substantial amounts of common stock in the public market, or the perception that such sales may occur, could adversely affect prevailing market prices.

Upon consummation of the offering, Healtheon will have an aggregate of 67,195,893 shares of common stock outstanding, based on the number of shares of common stock outstanding as of November 30, 1998, assuming that the U.S. underwriters do not exercise their over-allotment option and none of the outstanding options and warrants are exercised. Of the 67,195,893 shares outstanding after the offering, 5,658,184 shares, including the 5,000,000 shares sold in this offering, will be freely tradable without restriction under the Securities Act, except for any shares that may be purchased by "affiliates" of Healtheon. Shares purchased by Healtheon's affiliates will be subject to the volume and other limitations of Rule 144 of the Securities Act, or "Rule 144" described below. As defined in Rule 144, an "affiliate" of an issuer is a person who, directly or indirectly, through one or more intermediaries, controls, is controlled by or is under common control with the issuer. Upon the expiration of certain contractual "l ock-up" restrictions described below, 52,254,368 shares will be eligible for sale 180 days after the date of this prospectus, with 41,817,104 of such shares subject to the volume and other limitations of Rule 144. The remaining 9,283,341 shares will become eligible for sale at various times after that date, including 7,683,341 shares that will become eligible for resale between November 3 and November 6, 1999. All of these remaining shares will be subject to the volume and other limitations of Rule 144.

Each of Healtheon's directors and officers and certain other stockholders of Healtheon have agreed with Morgan Stanley & Co. Incorporated, for a period of 180 days after the date of this prospectus, not to:

-- offer, pledge, sell, contract to sell, sell any option or contract to purchase, purchase any option or contract to sell, grant any option, right or warrant to purchase, lend, or otherwise transfer or dispose of, directly or indirectly, any shares of common stock or any securities convertible into or exercisable or exchangeable for common stock; or

-- enter into any swap or other arrangement that transfers to another, in whole or in part, any of the economic consequences of ownership of the common stock, whether any such transaction described above is to be settled by delivery of common stock or other securities, in cash or otherwise.

Morgan Stanley & Co. Incorporated may choose to release some of these shares from such restrictions prior to the expiration of the 180-day period "lock-up" period, although it has no current intention of doing so.

Under Rule 144 as currently in effect, beginning 90 days after the date of this prospectus, a person who has beneficially owned restricted shares of common stock for at least one year, including the holding period of any prior owner who is not an affiliate, would be entitled to sell a number of the shares within any three-month period equal to the greater of(1) 1% of the then outstanding shares of the common stock or (2) the average weekly reported volume of trading of the common stock on the Nasdaq National Market during the four calendar weeks preceding such sale. Immediately after the offering, 1% of Healtheon's outstanding shares of common stock would equal approximately 671,959 shares. Under Rule 144, restricted shares are subject to manner of sale and notice requirements and requirements as to the availability of current public information concerning Healtheon. Under Rule 144(k), a person who is not deemed to have been an affiliate at any time during the 90 days preceding a sale, and who has beneficiall y owned the shares proposed to be sold for at least two years, including the holding period of any prior owner who is not an affiliate, is entitled to sell such shares without regard to the volume or other limitations of Rule 144 just described.

Immediately after this offering, there will be options to purchase approximately 11,827,385 shares of common stock outstanding, based on the number of options outstanding as of November 30, 1998. Subject to the provisions of the lock-up agreements described above, holders of these options may rely on the resale provisions of Rule 701 under the Securities Act. Rule 701 permits non-affiliates to sell their shares without having to comply with the volume, holding period, or other limitations of Rule 144 and permits affiliates to sell their shares without having to comply with the holding period limitation of Rule 144, in each case beginning 90 days after the consummation of this offering. In addition, shortly after this offering, Healtheon intends to file a registration statement on Form S-8 covering the 13,811,659 shares of common stock reserved for issuance under the 1996 Plan and the 1998 Purchase Plan based upon the number of options outstanding as of November 30, 1998. Shares of common stock registered unde r any registration statement will, subject to Rule 144 volume limitations applicable to affiliates, be available for sale in the open market, unless the shares are subject to vesting restrictions with Healtheon or the lock-up agreements described above.