The Performance of Internet Firms Following Their Initial Public Offering

Transcription

1 The Financial Review 37 (2002) The Performance of Internet Firms Following Their Initial Public Offering Jarrod Johnston University of Minnesota-Duluth Jeff Madura Florida Atlantic University Abstract We find that initial returns were more favorable for Internet initial public offerings (IPOs) than non-internet firm IPOs. Since the demise of the Internet sector, the underpricing of Internetfirm IPOs is not significantly different from other IPOs. Initial returns of Internet firms are positively and significantly related to underwriter prestige and to pre-ipo market conditions. However, initial returns after the demise of the Internet sector are not significantly related to these characteristics. The aftermarket performance of Internet firms is initially favorable but weakens over time. Firms that experienced higher initial returns during the strong Internet cycle experience weaker aftermarket performance. Keywords: IPO, initial public offering, underpricing, internet firms, cycles JEL Classifications: G14/G32 1. Introduction Internet firms are unique in that they offer much potential for future growth, which justifies their pursuit of substantial funding. The Internet has forced companies Corresponding author: Department of Finance, University of Minnesota Duluth, 150 SBE, 10 University Drive, Duluth, MN, ; Phone: (218) ; We thank Stephen P. Ferris (Editor), two anonymous reviewers, Ken Bartunek, Joel Harper, and Victor Kalafa for their helpful comments and suggestions. 525

2 526 J. Johnston and J. Madura/The Financial Review 37 (2002) in nearly every industry to reevaluate their business practices. Traditional channels of distribution are being eliminated, advertising methods are changing, supply chains are becoming more integrated, and consumers are able to compare prices easily and adjust their purchasing patterns accordingly. Yet, although the potential for the Internet is unlimited, the barriers to Internet entry tend to be low. The technological changes and the potential entry by new competition cause substantial uncertainty for firms that operate in the Internet environment. To the extent that the investors perceive that Internet firms are distinguished by a higher level of risk, underwriters could be forced to allow for additional underpricing at the time of issue. Ritter (1984) demonstrates that the popularity of initial public offerings (IPOs) can shift in response to a particular sector. He determines that the exceptionally high initial returns on IPOs in the early 1980s can be attributed to firms in the natural gas industry, which is evidence of a segmented market. This evidence is relevant to the study at hand, because it suggests that returns following an IPO can vary with the operations of the issuer. Because Internet firms have unique operations, they can exhibit a unique level of underwriting and therefore experience unique initial and aftermarket returns. Furthermore, they can be subject to shifts in investor sentiment. Because Internet firms exhibit a unique risk-return profile, they deserve to be assessed during and after their IPOs. Our objective is to determine the degree of underpricing of Internet firms at the time of the IPO and to determine whether the degree of underpricing is conditioned on characteristics peculiar to the firm or to the underwriter. A set of non-internet firms matched by industry, issue date, and size is created, so that the initial and aftermarket performance of Internet and non- Internet firms can be compared. The analysis contained in this study suggests that Internet IPOs generate significantly higher initial returns than non-internet IPOs. Moreover, the initial returns of Internet firms remain higher than non-internet firms after controlling for other firm-specific and issue-specific factors that could confound a direct comparison. The initial returns of Internet firms do not remain significantly higher after the demise of the Internet sector. Finally, the aftermarket performance of Internet firms initially exceeds that of non-internet firms, but the differential diminishes for longer time horizons beyond the IPO. 2. Underwriting and price stabilization process Underwriters facilitate IPOs with an underwriting and price stabilization process. Details about the underwriting and stabilization process are provided by Beatty and Ritter (1986), Benveniste and Spindt (1989), Hanley, Kumar, and Seguin (1993), and Michaely and Shaw (1994). In general terms, a firm that attempts to go public must prepare a detailed business plan and hire accountants, attorneys, and the investment bank that will serve as the lead underwriter. It must file documents with the Securities and Exchange Commission regarding its intent to offer securities to the public. These

3 J. Johnston and J. Madura/The Financial Review 37 (2002) documents must disclose financial data for the past five years, the composition and background of the board of directors, the level of competition, the outlook for the industry, and the intended use of the proceeds. The lead underwriter selected by the issuing firm creates a syndicate to place the shares. Syndicate members distribute the preliminary prospectus after its registration with the Securities Exchange Commission and solicit indications of interest from potential investors. These indications provide an early measure of the demand for the shares and influence the decision regarding the price and the number of shares to be offered. A final prospectus is created and distributed to investors. The lead underwriter and the firm s management meet to finalize the offer size and the offer price of the shares Underwriter goals The lead underwriting firm has multiple goals. It wants to set a price that is sufficiently high to satisfy the issuer; yet it also wants to set a price that is sufficiently low to ensure that all the shares are placed. Underwriters are motivated to underprice the IPOs so that the investors are satisfied, shares are placed, and investors can again participate in a seasoned offering by the issuer (see Welch, 1989). The goals of the underwriter when facilitating an IPO by an Internet firm are conceptually similar to the goals when facilitating an IPO by a non-internet firm. However, the characteristics of Internet firms can alter the behavior of the underwriter to satisfy those goals. Schultz and Zaman (2000) discuss motives by the issuer to go public, whereas Bartov, Mohanram, and Seethamarju (2001) attempt to identify the factors that drive the pricing of Internet IPOs Underwriter pricing Underwriters begin the pricing process by examining the past and projected financial statements of the offering firm and comparable firms in the industry. During the period prior to the offering, underwriters present the offer price range and solicit indications of interest from institutional investors. The indications of interest are nonbinding orders. Underwriters are able to adjust the price and number of shares offered according to the pre-offer demand, but the adjustments are normally limited. Once the offer price and size have been established, the declaration occurs, and the stock begins trading the next day. The potential growth of Internet operations combined with the potential ease of entry by future competitors makes the initial valuation of Internet firms very difficult for underwriters Relation between uncertainty and pricing Underwriters could be motivated to use a greater degree of underpricing when there is more uncertainty, because they are more likely to place the entire issue and the

4 528 J. Johnston and J. Madura/The Financial Review 37 (2002) price will be less likely to decline in the aftermarket. Beatty and Ritter (1986) suggest that initial returns are due to uncertainty. They state that informed investors will only invest in issues that are underpriced and oversubscribed. Their sample includes 1,028 firms from 1977 to Their proxies for ex-ante uncertainty are positively related to the initial return of IPOs over this period. Recent research offers some insight on the underpricing of Internet IPOs. DuCharme, Rajgopal, and Sefcik (2001) find that Internet underpricing is related to characteristics such as degree of media attention and the desire to pursue a secondary offering in the future. Ljungqvist, Nanda, and Singh (2001) specify a model that explains how high initial returns in a hot Internet IPO market can be attributed to a set of irrational investors. Ljungqvist and Wilhelm (2002) detect an abrupt shift in the initial returns of IPOs in the late 1990s, which they attribute to the Internet characteristics of the issuing firms. Van der Goot and Knauff (2001) find that the pricing of Internet IPOs in Europe is partially influenced by hot markets Underwriter efforts to stabilize prices Underwriters can attempt to ensure price stabilization by discouraging their investors from selling in the immediate aftermarket. In addition, underwriters can purchase shares if the price begins to fall. Underwriters frequently oversell issues if they have an overallotment option available. The overallotment option allows underwriters to obtain an additional specified percentage of shares at the offer price. Therefore, the underwriter oversells the issue and exercises the option to fulfill sales if the price rises in early aftermarket trading. If the price falls in early aftermarket trading, the underwriter can purchase shares in the market to stabilize the price and does not exercise the overallotment option. The existence and size of an overallotment option limits the downside potential by increasing demand. Underwriters can also prevent price reductions by imposing lock-up agreements, whereby investors are required to hold their shares for a specified amount of time. Lead underwriters can also influence their syndicate members through penalty bids. Portions of the fees earned can be reclaimed by the underwriter from syndicate members whose allocated shares are sold in early aftermarket trading. Faced with a potential loss of income, syndicate members inform their investors that those selling in the immediate aftermarket will not be allocated shares in future offerings. Both of these methods can artificially inflate the aftermarket price and contribute to high initial returns Research on stabilization efforts Some research has assessed price stabilization efforts. Ruud (1993) finds that one-fourth of IPOs have negligible returns and experience price declines following

5 J. Johnston and J. Madura/The Financial Review 37 (2002) Table 1 Review of findings of initial returns The initial return results from some notable past studies. Initial returns are the returns generated from the subscription price to the end of a short trading period following the beginning of trading. Author Time Period Initial Return (%) Ibbotson and Jaffe (1975) Reilly (1978) Ritter (1984) a lack of underwriter support. Hanley, Kumar, and Seguin (1993) finds that bid-ask spreads of dealers are smaller for the issues subject to stabilization. These issues also declined in value once the stabilization was discontinued. Schultz and Zaman (1994) find that underwriters repurchase large quantities of stock when the stock price is near or below the offer price in the first three days of trading. This provides further evidence that high initial returns could be partially attributed to the efforts of the underwriters to intervene for those issues in which the demand is weaker than expected. 3. Performance There is strong evidence of high initial returns following an IPO. Ritter (1984) finds average initial returns as high as 48.4% for IPOs that occurred during a 15-month period beginning in The mean return using a sample of more than 5,000 IPOs from 1960 to 1982 was 18.8%. Ritter demonstrates that initial returns fluctuate based on the time period and industry examined. A summary of studies that measured the initial returns is provided in Table Intraday returns Given that extraordinary abnormal returns occur by the end of the first day of trading, Barry and Jennings (1993) take a closer look at the first day. The first day is divided into an opening price return and an intraday return. The opening price return is calculated as the return earned from the offer price to the opening price, and the intraday return is calculated as the return earned from the opening price to the closing price. They find that 90% of the initial return occurred from the offer price to the opening price and therefore accrued only to subscribers. They conclude that the market quickly corrects the mispricing by the underwriters and only subscribers are able to earn high initial returns. Investors who are not allocated any shares do not earn abnormal initial returns by investing at the opening price.

6 530 J. Johnston and J. Madura/The Financial Review 37 (2002) Aftermarket performance Aggarwal and Rivoli (1990) suggest that IPOs are valued incorrectly only after they begin trading in the market. They examine the returns for one year following issuance exclusive of the first trading day and found that the return is significantly less than the market. Ritter (1991) compares the three-year returns of firms that engaged in IPOs from 1975 through 1984 to benchmarks and a set of matching firms. He finds that the IPOs significantly underperform both market benchmarks and matching firms in the aftermarket. Loughran and Ritter (1995) also address the issue of aftermarket performance. Using a sample of 4,753 companies going public from 1970 to 1990, they find weak performance over a period of five years beyond the offering. Aggarwal and Rivoli (1990) explain that mispricing occurs in early trading of IPOs when investors drive up the price to irrational levels. This explanation is relevant for Internet IPOs because it suggests that Internet firms could experience a correction when the trading just after the IPO drives prices to irrational levels. The mispricing could occur at the time of the offer, if underwriters underprice beyond what is appropriate based on the uncertainty. Alternatively, it could occur shortly after the offering, especially when the stock becomes accessible to uninformed individual investors. 4. Hypotheses Given the unique valuation characteristics of Internet firms, the following hypotheses are developed Unique initial returns of Internet firms Accounting variables have shown limited ability to assist in the valuation of IPOs. Underwriters frequently use accounting information about firm-specific characteristics to price IPOs. Kim and Ritter (1999) compare historical earnings, marketto-book, price-to-sales, and other accounting ratios of new offerings to comparable existing firms to determine an appropriate price range. They find limited predictive ability, which is attributed to wide variation in these ratios. Moreover, to the extent that Internet firms exhibit unique operations, their financial characteristics are not necessarily comparable to other firms. Thus, the valuation of these firms can be subject to greater error and therefore a greater degree of underpricing. Ritter (1984) and others have demonstrated that initial returns are directly related to the uncertainty of the IPO. Underwriters must underprice to entice investment into uncertain offerings. This underpricing must increase with the uncertainty to ensure that the offering is fully subscribed. The high degree of underpricing is reflected in abnormally high initial returns when compared to similar firms that do not focus their operations around the Internet. Therefore, the initial returns of firms engaging in an IPO and operating

7 J. Johnston and J. Madura/The Financial Review 37 (2002) primarily through the Internet should be significantly higher than those of non-internet firms Underwriter certification of Internet firms Extensive research has been conducted on the relation between underwriter prestige and initial returns of IPOs. Beatty and Ritter (1986) suggest that initial returns of IPOs are conditioned on the underwriter, as prestigious underwriters are less likely to underprice their offerings. Johnson and Miller (1988) and Carter and Manaster (1990) find that underpricing is inversely related to underwriter reputation. They also found that prestigious underwriters are involved with less risky issues. Beatty and Ritter (1986) determine that prestigious underwriters are more likely to fully price their offerings. They also show that underwriters that overprice lose market share in future periods. Carter and Dark (1992) expand on the importance of underwriter reputation by investigating the best method to measure reputation. A sample of 1,267 offerings is used to examine the Carter and Manaster method and Johnson and Miller s modified Carter-Manaster method. They regress the initial return on the standard deviation of after-market return (a proxy for risk) and the underwriter reputation. The model with the modified Carter-Manaster proxy for underwriter reputation has a significantly positive coefficient on the risk proxy and an insignificant coefficient on the coefficient for underwriter reputation. The model with the Carter and Manaster proxy for underwriter reputation has significantly positive coefficients on both risk and underwriter reputation. A test of both models indicates that only the Carter and Manaster measure is significant. Michaely and Shaw (1994) reexamine the issue using the investment bank s capital as the proxy for prestige. Twenty-three underwriters accounted for 50% of the transactions and are classified as prestigious. They find that IPOs of prestigious underwriters had significantly lower initial returns, which implies a relatively low level of underpricing. Dunbar (2000) measures the impact of underpricing on the subsequent IPO market share of underwriters. The initial return is inversely related to the subsequent market share of reputable investment banks. As Internet firms became popular, the prestigious underwriters did not shun them. In fact, 137 of the 366 Internet firms in the sample were underwritten by the most prestigious underwriters when applying the Carter and Manaster ranking (those underwriters with a score of nine). Thus, the typical behavior of prestigious underwriters to avoid very risky offerings (see Carter and Manaster, 1990) did not apply during the advent of Internet-firm IPOs. The Internet firms could have pursued prestigious underwriters to achieve credibility, because the pre-ipo performance of these firms was not by itself convincing to prospective investors. Some of these offerings were very large and attractive to prestigious underwriters, given that the fees are typically assessed as a percentage of the issue. Had a prestigious underwriting

8 532 J. Johnston and J. Madura/The Financial Review 37 (2002) firm avoided Internet-firm IPOs, it would have forgone opportunities that would be pursued by other prestigious underwriters. Because prestigious underwriters are involved in underwriting numerous IPOs of Internet firms, the relation between underwriter prestige and initial return deserves to be reassessed. To the extent that prestigious underwriters are focused more on Internet firms than non-internet firms, initial returns can be positively associated with underwriter prestige. That is, if the market is concerned about the high valuations of Internet firms that are subject to much uncertainty, it could reward those Internet IPOs that have the support of more credible underwriters Marketing impact on initial returns Several firms with Internet related operations include dotcom in their name, which serves to immediately signal the firm s focus on Internet operations. This alone could increase initial returns and indicates either investor irrationality in early aftermarket trading or that different marketing strategies can be very effective. Therefore, Internet firms with dotcom in their name are expected to have higher initial returns Aftermarket performance To the extent that underpricing is not fully absorbed by the initial returns, it can allow for higher returns in the aftermarket. In addition, the investor optimism that drives initial returns will be extended to the months following the offering. This risk and enthusiasm will be reflected in positive aftermarket returns. However, because the Internet sector is subject to abrupt shifts in investor sentiment, the aftermarket effects are expected to vary with investor sentiment. 5. Research design The Internet is not an industry but simply an instrument that companies from a wide variety of businesses can use. Booksellers, automobile dealerships, and music stores are just a few examples of industries that have firms focusing on the Internet. Therefore, it is possible to have public offerings of companies that are in the same industry, with some planning on using the Internet as a prime source of revenue and others using traditional retail outlets. This situation provides the opportunity to compare the initial returns of two companies in the same industry going public at nearly the same time. Our sample of 366 Internet-related IPOs from 1996 through 2000 was compiled from Edgar Online and Hoover s IPO Central. Firms included in the sample are those that indicated their operations will be dependent on the Internet. A set of matching firms was also compiled from the Securities Data Corporation s Database of Global New Issues. The matching sample of firms without Internet operations was selected by industry, date, and size, respectively.

9 J. Johnston and J. Madura/The Financial Review 37 (2002) A company that sells products through the Internet is matched with a traditional retailer. Software companies that focus their business over the Internet are matched with software companies that did not emphasize the Internet, although Internet operations could exist. Most companies have an Internet presence. To be classified as an Internet company, it had to be critical to their operations. The following excerpts are from the company descriptions in the registration filing with the Securities Exchange Commission. The first is classified as an Internet company, and the second is used as a matching firm. Allaire develops, markets, and supports application development and server software for a wide range of Web development, from building static Web pages to developing enterprise-scale, interactive Web applications. The Company s products and services enable professional Web developers to build high-volume, transaction-oriented Web sites and 0 applications for electronic commerce, business information systems, content publishing, and multi-user collaboration. Actuate Software Corporation is a leading provider of enterprise-reporting solutions that enable organizations to systematically extract, publish, and disseminate information across distributed computing environments. The Company develops and markets software products that are designed to allow companies to rapidly design, generate, and distribute reports throughout the enterprise, thereby increasing access to and the value of corporate data. Similar to Ritter (1984), the initial return is calculated from the offering price to the close of the first day of trading and is not market- or matching-adjusted. Aftermarket returns are measured as buy-and-hold returns for the specified time period and are exclusive of the initial return. Similar to Brav and Gompers (1997), wealth relatives are calculated using the Nasdaq Composite Index and the matching firms as benchmarks. Wealth relatives are computed by (1 + R i,t )/ (1 + R bench,t ), where R i,t is the buy-and-hold return for IPO i for period t. R bench,t is the buy-andhold return for the benchmark for period t. Given the short time since Internet IPOs occurred, the aftermarket assessed here is substantially shorter than the aftermarket horizon used by Loughran and Ritter (1995) to assess aftermarket effects following offerings. 6. Descriptive statistics of sample Descriptive statistics are provided in Table 2 for Internet firms (Panel A) and matching firms (Panel B). The size of the offerings and the number of firms with dotcom in their name increased with the number of IPOs. Only 8 Internet firms went public in 1996, whereas 224 went public in Of the total IPOs in 1999, Internet firms made up more than 50%. After heavy IPO volume by Internet firms in the first quarter of 2000, the Internet sector experienced a dramatic downturn. IPOs by Internet firms declined after that point and made up only 27.58% of all IPOs in As Table 2 shows, the days in registration and mid-file price varies with the sample period. The offer price generally increased during the period.

10 Table 2 Descriptive statistics for Internet IPOs Descriptive statistics for the sample of firms with Internet operations and the sample of matching firms that went public from January 1, 1996, to December 30, The variables include the number of firms, the number of IPOs as a percentage of the total number of IPOs, the average amount raised, the average asset size prior to going public, the number of firms with dotcom in their name, the average days in registration, the average level of inside ownership, the number of firms with venture capital investment, the number of firms going public on the Amex, the NYSE, and the NASDAQ stock exchange, the average mid-file price, and the average offer price. Panel A: Firms with Internet operations Percent of Days in Inside Mid-File Offer Time N Total IPOs Mean Offer Mean Assets Dotcom Registration Ownership Venture Amex NYSE NASDAQ Price Price ,887,500 8,550, ,970,000 28,930, ,417,391 62,134, ,540, ,444, ,196,703 77,048, ,901, ,008, Panel B: Matching firms ,600,000 69,954, ,163,889 61,455, ,340, ,685, ,102, ,000, ,658,716 80,743, ,690, ,058, J. Johnston and J. Madura/The Financial Review 37 (2002)

11 J. Johnston and J. Madura/The Financial Review 37 (2002) Table 3 Means tests of initial returns The sample time period is January 1, 1996, to December 31, The average percent initial return is calculated for the sample of Internet firms, a subset of dotcom firms, and the sample matching firms. Panel A: Average initial returns and t-test that the initial return is zero Sample N Average Initial Return t-statistic Internet firms Dotcom firms Matching firms Panel B: Two-Tailed t-test that the Initial Returns Between the Internet Firms and Matching Firms are Equal Sample Mean Difference t-statistic Internet and matching firms Indicates statistical significance at the 0.01 level. The size of the offering and days in registration for the Internet firms are similar to matching firms. The insider ownership is much higher for Internet IPOs than matching IPOs. More of the Internet firms were backed by venture capital than matching firms. Most of the Internet and matching firms that engaged in IPOs over the 1996 to 2000 period are listed on the Nasdaq market. The mid-file price is slightly higher for matching firms, whereas the offer price is slightly higher for Internet firms. 7. Measurement of initial returns The hypothesis regarding the initial returns of Internet firms versus non-internet firms is tested by comparing the initial returns of Internet firms at the time of their respective IPOs to a matched set of non-internet firms. A paired-sample t-test is used to measure the significance of the difference between the Internet firms and their matching counterparts. As shown in Table 3, Internet firms experience higher initial returns, and the difference is significant at the.01 level. The higher initial returns of Internet firms are attributed to a greater amount of uncertainty about their future performance, which requires a greater degree of underpricing by underwriters. The results could be attributed to other characteristics about the offering that are not captured when simply comparing the Internet firms with the matching firms. For this reason, a multivariate model is also applied to control for other characteristics. 8. Cross-sectional analysis of initial returns A cross-sectional analysis of initial returns can determine whether the initial returns are higher for Internet firms than non-internet firms, while controlling for other characteristics that could affect the initial returns.

12 536 J. Johnston and J. Madura/The Financial Review 37 (2002) Internet dummy An Internet dummy variable is included in the model to determine whether initial returns among firms in the sample to test are higher for Internet firms. To distinguish Internet firms from non-internet firms, a dummy variable is assigned a value of 1.0 for Internet firms and zero otherwise. The remaining characteristics are classified as issue-related or market-related Issue-related characteristics Underwriter rank Underwriter prestige could influence the initial returns (Carter and Manaster, 1990). A greater degree of underwriting is normally associated with less prestigious underwriters, which could be attributed to more uncertainty surrounding the offering. Carter and Manaster suggested that prestigious underwriters are less likely to underwrite risky offerings and could be more skilled at pricing their offerings appropriately. The underwriter prestige variable (RANK) is measured by applying the Carter-Manaster ranking method from Carter, Dark, and Singh (1998) Overallotment An overallotment option allows the underwriter to sell more shares of an offering than the underwriter has pledged to sell. When there is excess aftermarket demand, the underwriter simply exercises the overallotment option to gain the shares needed to fill the sales. If the price falls in the immediate aftermarket, the underwriter will simply purchase shares in the market to fill its sales. Generally, the underwriter has the option to purchase an additional 15% of the shares raised in the IPO, which allows overselling of the offering. The percentage of shares included as part of an overallotment option could be directly related to the initial return. The initial return of Internet firms engaging in an IPO in which the underwriter is using an overallotment option as a price stabilizer is expected to be significantly higher than Internet firms in which the underwriter is not using these price stabilization techniques. This variable (OLT) is the percentage amount of overallotment exercised Change in price from filing to offering The type of book-building efforts made by the underwriter can influence the demand for IPO shares. A relatively large increase in the price from the filing to the offer could suggest that the indications of interest in the IPO are stronger than expected. Consequently, the initial return is expected to be relatively high for offerings in which the price change from the filing to the offering was high. This variable (PCHG) is measured as the percentage change from the mid-file price to the offer price. Additionally, the asymmetric effect of the change is included in the model (PCHGA). This variable is equal to PCHG when positive and zero otherwise. Lowry

13 J. Johnston and J. Madura/The Financial Review 37 (2002) and Schwert (2002) find that the price change and asymmetric changes have an effect on initial returns Offer size Larger offerings attract more scrutiny from analysts and are issued by larger, more recognizable firms. These characteristics can reduce uncertainty and limit initial returns. The log of the gross proceeds raised in the IPO (PROCEEDS) and the log of the firm s assets prior to the IPO (ASSETS) are used to measure offer size Lockup agreements Lockup agreements prohibit insiders from selling shares for a specified time period, typically six months. These agreements restrict the supply of shares that can be sold by insiders (see Keasler, 2001) and could increase the initial return. The variable (LOCK) is measured as the number of shares subject to the lockup agreement divided by the number of shares offered to the public Venture capital Venture capitalists could potentially offer expertise to firms that are about to go public. Thus, there could less uncertainty surrounding these firms, which could allow for a smaller amount of underpricing (lower initial returns). Conversely, if IPOs backed by venture capital attract stronger demand that is not fully captured by the offer price, initial returns could be higher. The variable (VENTURE) is an indicator variable equal to one if there is venture capital investment and zero otherwise. An interaction variable (INTVENT) is included in the model to determine the additional effect that venture capital has on Internet firms. It is calculated by multiplying the Internet variable and the venture capital variable Inside ownership The level of inside ownership could provide an indication of the uncertainty of the offering. Higher levels should reduce uncertainty and reduce initial returns. However, to the extent that insiders could be motivated to liquidate their holdings (see Keasler, 2001). The variable (INOWN) is measured as the percent of total shares outstanding held by insiders Market-related characteristics Pre-IPO market conditions The initial return could vary with the mood of the market at the time of the IPO. More favorable pre-ipo market conditions can cause more optimistic expectations about the prospects of the firm and a more favorable initial return. The pre-ipo market

14 538 J. Johnston and J. Madura/The Financial Review 37 (2002) conditions (RUNUP) are measured as the return on the Nasdaq market over the twoweek period preceding the IPO. Additionally, the asymmetric effect of the change is included in the model (RUNUPA). This variable is equal to RUNUP when positive and zero otherwise. Lowry and Schwert (2002) find that the market conditions and asymmetric changes have an effect on initial returns Market downturn The market, particularly for Internet firms, declined significantly at the end of March The number of firms going public also declined as the market declined. The initial returns associated with IPOs could be affected by the abrupt weakening of the market and shift in sentiment away from Internet stocks. An indicator variable is used to measure the market downturn (DOWN), which is equal to one if the offer date is after March 31, 2000 and zero otherwise. Additionally, the downturn could have a more pronounced effect on Internet firms going public. An interaction variable (INTDOWN) is included in the model to measure this effect, which is computed as the Internet variable multiplied by the downturn variable Stock exchange of offering Reporting and capitalization requirements differ between stock exchanges and could have an effect on how an IPO is perceived. Indicator variables are used for the Amex and the NYSE to determine whether initial returns are conditioned on where the stock of the firm going public is listed Model When modeling the cross-sectional variation in initial returns among firms that engaged in IPOs, estimation errors can occur due to omitted variables. Related research suggests a wide variety of characteristics that can be related to initial returns. Therefore, we first use a model that contains all characteristics that could be measured and can help to explain the variation in initial returns. The following model (referred to as the complete model) is used to explain the variation in initial returns among the pooled sample of Internet and non-internet firm IPOs. 1 IR = β 0 + β 1 INTERNET + β 2 RANK + β 3 OLT + β 4 PCHG + β 5 PCHGA + β 6 PROCEEDS + β 7 ASSETS + β 8 LOCK + β 9 VENTURE + β 10 INTVENT + β 11 INOWN + β 12 RUNUP + β 13 RUNUPA + β 14 DOWN + β 15 INTDOWN + β 16 AMEX + β 17 NYSE + ε (1) 1 Three pairs of variables exhibited a high degree of collinearity. PCHG and PCHGA, PROCEEDS and ASSETS, and RUNUP and RUNUPA were orthogonalized to remove the correlation. This transformation also applies to the cross-sectional analysis of the aftermarket returns.

15 J. Johnston and J. Madura/The Financial Review 37 (2002) A restricted model that excludes some of the variables is also applied to test the entire period, as well as the subperiods representing strong and weak cycles for Internet stocks Application of complete model to the entire period The results for the entire period are disclosed in Table 4. The results are shown for four different applications. The complete model that contains all variables described previously is applied to the entire period (column 1). A reduced-form model is also applied to the same period, with the results reported in column 2. That model is also applied to the strong cycle for Internet stocks (up to March 2000) and to the weak Internet stock cycle (after March 2000). These results are summarized in columns 3 and 4 and are described later. For the entire period, the coefficient on the INTERNET variable is positive and significant, which indicates superior initial returns of Internet firms even when controlling for other factors. The results are attributed to the high degree of uncertainty inherent in Internet firms, which results in substantial underpricing. These results support the hypothesis of higher initial returns for Internet firms. The OLT variable is positive and significant, which supports the hypothesis that the initial returns are more favorable when the amount of overallotment is larger (when the demand conditions are stronger). However, this does not mean that overallotment is the cause of the higher initial return. It is quite possible that the initial return would have been higher if the lead underwriter did not exercise some overallotment and that the price could have drifted downward in the aftermarket as a correction. The two variables representing the change from the mid-file price to the offer price (PCHG and PCHGA) are positive and significant. In general, the underwriter adjusts the offer price in accordance with the demand indications. However, the offer price is not fully adjusted, as it is associated with higher levels of demand and therefore higher initial returns. The VENTURE variable is positive and significant, implying a stronger initial return for IPOs backed with venture capital. The RUNUP variable is positive and significant, which supports the hypothesis that IPOs have higher initial returns when preceded by strong pre-issue market conditions. The other variables are not statistically significant when applied to the entire period Application of reduced-form model to the entire period The entire period is reassessed with a reduced-form model, which excludes some of the variables that are not significant when applying the complete model. The INTERNET variable remains positive and significant. The RANK variable is positive as it was before but is now significant at the.10 level. This relation suggests that during the period in question, prestigious underwriters commonly served as underwriters for risky stocks (including Internet stocks). They underpriced these IPOs to reflect the

16 Table 4 Cross-sectional regression effects related to Internet and matching firm IPO initial returns The sample of firms with Internet operations and matching firms that went public from January 1, 1996, to December 31, The dependent variable is the initial return. INTERNET is an indicator variable equal to one if the company had Internet operations and zero otherwise. RANK is the level of underwriter prestige from Carter, Dark, and Singh (1998). OLT is the percent of shares exercised in an overallotment agreement. PCHG is the percentage change from the mid-file price to the offer price. PCHGA is equal to PCHG when it is positive and zero otherwise. PROCEEDS is the log of proceeds raised in the IPO. ASSETS is the log of assets of the firm prior to the IPO. LOCK is the percent of shares subject to a lockup agreement. VENTURE is an indicator variable equal to one if there is venture capital investment and zero otherwise. INTVENT is an interaction variable between INTERNET and VENTURE. INOWN is the percentage of inside ownership prior to the IPO. RUNUP is the return on the Nasdaq Composite Index for the two weeks prior to the IPO. RUNUPA is equal to RUNUP when it is positive and zero otherwise. DOWN is an indicator variable equal to one if the offer date is following March 31, 2000, and zero if the offer date is prior to that date. INTDOWN is an interaction variable between INTERNET and DOWN. Amex is an indicator variable equal to one if the firm came to market on the Amex and zero otherwise. NYSE is an indicator variable equal to one if the firm came to market on the NYSE and zero otherwise. Jan. 1, 1996 Dec. 31, 2000 Jan. 1, 1996 Dec. 31, 2000 Jan. 1, 1996 March 31, 2000 March 31, 2000 Dec. 31, 2000 Variable Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Intercept INTERNET RANK OLT PCHG PCHGA PROCEEDS ASSETS LOCK VENTURE INTVENT INOWN RUNUP RUNUPA DOWN INTDOWN Amex NYSE Adj. R F value N Indicates statistical significance at the 0.01 level. Indicates statistical significance at the 0.05 level. Indicates statistical significance at the 0.10 level. 540 J. Johnston and J. Madura/The Financial Review 37 (2002)

17 J. Johnston and J. Madura/The Financial Review 37 (2002) uncertainty. In other words, investors rewarded firms whose stock was underwritten by more credible underwriters. The remaining results when applying the reduced-form model are similar to those derived from the complete model. All other variables that are significant when using the reduced-form model are significant when using the complete model, and the sign of each coefficient is not affected by the model used Application of the reduced-form model to the strong and weak Internet cycles The restricted model is applied to a subperiod representing a strong Internet cycle and a subperiod representing a weak Internet cycle. 2 Internet firms that came to market after the demise of the Internet sector were forced to offer more credible business models to justify their request for funding. With more substance backing the IPO, investors could rely more on information about the firm itself rather than other characteristics when deciding whether to participate. To the extent that the sensitivity of investor participation in IPOs is affected by the shift in the valuation process since the demise of the Internet sector, the sensitivity of initial returns of IPOs to crosssectional characteristics could have changed as well. Therefore, the reduced-form model is separately applied to two subperiods. 3 Results from applying the reduced-form model to each subperiod are shown in the last two columns of Table 4. Thirty-three Internet IPOs and their respective matching IPOs were deleted because they straddled the March 31 break point between the two subperiods. The INTERNET variable is positive and significant when applied to the strong Internet cycle, which is consistent with the results found for the entire period. However, the INTERNET variable is not significant when applied to the weak Internet cycle, which suggests that the higher degree of underpricing for Internet stocks vanishes since the demise of the Internet sector. One possible explanation is that since the demise of the Internet sector, Internet firms that came to market needed a more credible business model that was more transparent to investors. The number of Internet firms that engaged in IPOs declined abruptly after the first quarter of Those that did come to market had to pass a more stringent screening by the market and exhibited no more uncertainty than non-internet firms that went public at 2 See The Fall of the Net Analyst, Business Week, December 11, 2000, pp Ownership and oversight by insiders and venture capitalists could reduce the amount of uncertainty and increase investor demand for an IPO by Internet firms. Inside ownership indicates entrepreneurial confidence in the firm, whereas venture-backed IPOs have been found to outperform other IPOs (see Brav and Gompers, 1997). Inside ownership is measured as a percentage of total shares. The venture capital backing is measured as an indicator variable set to one if there is investment and zero otherwise. Neither variable is found to be significant when applied to Internet firms. The other variables identified in Footnote 1 were also tested for the sample of Internet firms and were not significant.

18 542 J. Johnston and J. Madura/The Financial Review 37 (2002) that time. Thus, these Internet firms that came to market were underpriced at a level similar to other types of firms. For the strong Internet stock cycle, the RANK variable has the same coefficient as it did when applied to the entire period. However, its standard error is slightly larger in the strong cycle period, which causes the t-statistic to fall short of significance. When applied to the weak Internet cycle, this variable has a much smaller coefficient and is far from the critical level of significance. The OLT variable is positive and significant during the strong and weak Internet cycles, consistent with the results found for the entire period. Internet firms experience higher initial returns when the underwriter uses price stabilization irrespective of the cycle. The variables representing the change from the mid-file price to the offer price are positive and significant in the strong and weak Internet cycles, consistent with the results found for the entire period. The VENTURE variable is positive and significant when applied to the strong Internet cycle but not significant when applied to the weak Internet cycle. It appears that the initial returns since the demise of the Internet sector no longer are associated with whether the IPO has some venture capital support. The RUNUP variable is also positive and significant in the strong Internet cycle but not in the weak Internet cycle. IPOs had higher initial returns when preceded by strong pre-issue market conditions when stock market conditions were favorable but not after the demise of the Internet stocks. 9. Variation in initial returns of Internet firms 9.1. Application of the complete model to the entire period Because the sensitivity of initial returns to characteristics differs for Internet firms, an additional cross-sectional analysis is conducted using the following model. The sample is limited to Internet firms only. The complete model is applied to this special sample over the entire period, with the following changes. First, because the sample is restricted to Internet firms only, the INTERNET variable and two interaction terms are not included in the model. Second, an indicator variable (DOTCOM) is included to capture the effect of the dotcom factor. The variable is equal to one if the firm has dotcom in its name and zero otherwise. IR = β 0 + β 1 DOTCOM + β 2 RANK + β 3 OLT + β 4 PCHG + β 5 PCHGA + β 6 PROCEEDS + β 7 ASSETS + β 8 LOCKUP + β 9 VENTURE + β 10 INOWN + β 11 RUNUP + β 12 RUNUPA + β 13 DOWN + β 14 AMEX + β 15 NYSE + ε (2) A reduced-form model is also applied. Results are shown in the second column of Table 5. The coefficient of the DOTCOM variable is not significant. Thus, there is no evidence that dotcom-firm IPOs have higher initial returns than other Internet-firm IPOs. The RANK variable is positive and significant, implying that initial returns are higher when underwritten by prestigious underwriters. The VENTURE variable is

19 Table 5 Cross-sectional regression effects related to Internet-firm IPO initial returns The sample of firms with Internet operations that went public from January 1, 1996 to December 31, The dependent variable is the initial return. DOTCOM is an indicator variable equal to one if the company had dotcom in its name and zero otherwise. RANK is the level of underwriter prestige from Carter, Dark, and Singh (1998). OLT is the percent of shares exercised in an overallotment agreement. PCHG is the percentage change from the mid-file price to the offer price. PCHGA is equal to PCHG when it is positive and zero otherwise and orthogonalized. PROCEEDS is the log of proceeds raised in the IPO. ASSETS is the log of assets of the firm prior to the IPO, orthogonalized with PROCEEDS. LOCK is the percent of shares subject to a lockup agreement. VENTURE is an indicator variable equal to one if there is venture capital investment and zero otherwise. RUNUP is the return on the Nasdaq Composite Index for the two weeks prior to the IPO. RUNUPA is equal to RUNUP when it is positive and zero otherwise and orthogonalized. DOWN is an indicator variable equal to one if the offer date is following March 31, 2000 and zero if the offer date is prior to that date. INOWN is the percentage of inside ownership prior to the IPO. Amex is an indicator variable equal to one if the firm came to market on the American Stock Exchange and zero otherwise. NYSE is an indicator variable equal to one if the firm came to market on the NYSE and zero otherwise. Jan. 1, 1996 Dec. 31, 2000 Jan. 1, 1996 Dec. 31, 2000 Jan. 1, 1996 March 31, 2000 March 31, 2000 Dec. 31, 2000 Variable Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Intercept DOTCOM RANK OLT PCHG PCHGA PROCEEDS ASSETS LOCKUP VENTURE INOWN RUNUP RUNPA DOWN Amex NYSE Adj. R F value N Indicates statistical significance at the 0.01 level. Indicates statistical significance at the 0.05 level. Indicates statistical significance at the 0.10 level. J. Johnston and J. Madura/The Financial Review 37 (2002)