Printer Friendly

Who makes the choice on IPO underwriting methods? Issuers versus underwriters.

Using unique Taiwanese initial public offering (IPO) features, we examine whether the underwriters control the issue method choice (the expected revenue collection hypothesis) or whether the issuers make the choice in the best interests of the pre-IPO shareholders (the expected net wealth gain hypothesis). We find that the choice of issue method is consistent with maximizing pre-IPO shareholders ' expected net wealth gains. The findings also suggest that most IPO issuers do not choose auctions. Though we offer insights into why IPO auctions have been declining, we note that this evidence cannot completely explain the total disappearance of auctions in many countries.

**********

In an initial public offering (IPO), the issuer or the lead underwriter has to choose an underwriting contract (firm commitment vs. best efforts) and/or an issue method (book-building, auction, or fixed price public offer). These choices affect both the issuer's realized offering costs and the underwriter's realized revenues. However, a relatively unexplored research question is: who makes the choice in the negotiation between an issuing firm and a leading underwriter?

The extant literature does not examine who plays a more important role in choosing an IPO underwriting contract or an issue method. Some studies assume that IPO issuers choose underwriting contracts (Mandelker and Raviv, 1977; Bower, 1989; Dunbar, 1998). In choosing an IPO issue method, Jagannathan, Jirnyi, and Sherman (2015) suggest that issuers avoid IPO auctions due to the complexity of the process, even though auctions have been used in more than 25 countries. In contrast, underwriters may dominate the IPO process decision and choose an issue method that maximizes their revenues because of better information and/or bargaining power (Ritter, 1984; Ausubel, 2002). In this paper, we explore the question by analyzing an IPO sample consisting of both fixed price public offerings and hybrid auction offerings in Taiwan from 1996 to 2003.

The market structure of IPO underwriting in Taiwan differs from that in the United States. Degeorge, Derrien, and Womack (2010) discuss US IPO auctions lead-managed by WR Hambrecht from 1999 to 2007. During this period, almost all of the other underwriters used book-building as their primary underwriting method. (1) In contrast, in our sample period from 1996 to 2003, 20 underwriters (of the top 42 largest underwriters in Taiwan) used both auctions and fixed price public offers. Therefore, this is an appropriate sample to study whether the IPO issuer or the underwriter made the decision in the choice of issue methods as both parties may have had different incentives. Furthermore, the IPO market in Taiwan provides a natural experiment to explore the choice between fixed price public offers and hybrid auctions. Such an experiment is not feasible in other countries as many other countries that have used auctions dropped the underwriting method of auction relatively quickly. In contrast, both the hybrid auction and the fixed price public offer have been available in Taiwan since 1995. Thus, the existence of both underwriting mechanisms for a long period of time makes the IPO auction sample in Taiwan relatively unique.

We propose two hypotheses in our paper. The first focuses on the IPO issuer's expected net wealth gain, while the second focuses on expected underwriter revenues to explain who makes the choice regarding the issue method. Specifically, the expected net wealth gain hypothesis posits that it is more likely for an IPO issuer to choose a specific issue method if the expected net wealth gain received via this method is larger than the gain obtained from choosing an alternative issue method. In contrast, the expected revenue collection hypothesis posits that it is more likely for the lead underwriter of an IPO to choose a specific issue method if the expected underwriter's revenues received when choosing this method are larger than the expected revenues earned from choosing an alternative issue method.

Beginning with the expected net wealth gain hypothesis, we compute the net wealth gain difference as the net wealth gain of pre-IPO shareholders if the IPO firm selects the fixed price public offer, minus the net wealth gain, if the firm chooses the hybrid auction. We find that 83% (519 of 623 IPOs) of IPO firms that chose a fixed price public offer would be worse off if they had chosen a hybrid auction instead. In contrast, only 24% (21 of 89 IPOs) of the IPO issuers that selected a hybrid auction would be worse off if they had chosen a fixed price public offer instead. Overall, among 712 IPO firms, 587 firms (82%) are or would have been better off when choosing a fixed price public offer, indicating that most IPO issuers in Taiwan would be better off choosing a fixed price public offer in terms of the expected net wealth gain. The results suggest that most issuing firms have significant negotiation power in choosing the underwriting method.

Next, turning to the expected revenue collection hypothesis, we note that in addition to underwriting fees, underwriters in Taiwan also earn subscription fees (also called lottery processing fees) from investor subscriptions to IPO shares. Chen, Fok, and Wang (2006) find that the average IPO underwriting fee is 0.99% and the average lottery processing fee is 5.62% of IPO proceeds from 1989 to 1999. Lottery processing fees accounted for a large proportion of the total revenues to underwriters. (2) In our sample period, the available number of shares in the fixed price tranche decreases to 50% if the hybrid auction procedure is selected as the other 50% of IPO shares is first offered in the auction tranche. Thus, we posit that underwriters encourage issuing firms to adopt the fixed price public offer so that underwriters can earn more revenue from the lottery processing fees. However, our results do not support the expected revenue collection hypothesis.

Overall, this paper relates to several strands of the IPO literature. First, we provide evidence to answer an important question in IPO decision-making. Who makes the choice between different IPO issue methods? We find that IPO issuers appear to have more bargaining power in choosing an issue method than underwriters do.

In addition, our paper links to the IPO auction literature. The evidence regarding who chooses the issue method may play a role in examining why IPO auctions have not been more popular around the world. Is it because underwriters have suppressed them, or because issuers chose to not use them? By deriving and examining the expected net wealth gain, we find that most issuers in Taiwan chose to not use IPO auctions aggressively.

While our study sheds some light on the "IPO auction demise" phenomenon by providing evidence that most issuers decline the use of IPO auctions, caution is still needed for interpretation. Since a few issuers in our sample were better off when using the hybrid auction, this evidence cannot completely explain the total disappearance of auctions.

Finally, when comparing issue methods, it is important to explore beyond the choice of the offer price and consider the possible effects of the method on the aftermarket price as firms are also concerned with the price of their stock in the secondary market for several reasons. Many firms seek to issue additional equity in the years following their IPOs. In addition, if the secondary market price continues to be relatively low, firms are vulnerable to a takeover at an undervalued price (Chemmanur and Liu, 2006). Moreover, firm insiders wish to sell their shares to diversify their portfolios or to meet their liquidity needs in the secondary market. This is also an important motive in the decision to go public (Pagano, 1993; Pagano, Panetta, and Zingales, 1998; Chen, Chen, and Huang, 2012).

In analyzing firms' choices between fixed price public offers and the auction method, Chemmanur and Liu (2006) present a model illustrating that firm insiders consider the extent of information production by outsiders. This information is reflected in the secondary market price, yielding a higher secondary market price for higher intrinsic-value firms. In their model, firms wish to obtain higher proceeds from the sale of stock, but they also care about the secondary market price of their stock following their IPOs. Similarly, the aftermarket price is one key component in calculating the expected net wealth gain of the pre-IPO shareholders in our study.

Moreover, the aftermarket price may depend upon how the IPO is conducted. Liu, Sherman, and Zhang (2014) find that firms with more pre-IPO media coverage have higher long-term stock prices relative to the fundamentals, as well as greater liquidity, analyst coverage, and institutional investor ownership in the three years following the IPO. They use an instrumental variable approach to illustrate that shifts in pre-IPO attention affect the aftermarket stock price. The choice of issue method in Taiwan may affect the investor attention that an IPO firm receives in the aftermarket, thus explaining why the aftermarket price may be dependent upon the issue method used.

The rest of this paper proceeds as follows. Section I reviews the literature regarding the choice in IPO underwriting methods. Section II discusses fixed price public offers and hybrid auctions in Taiwan. Section III develops the hypotheses to explore the choice between fixed price public offers and hybrid auctions. It also presents the methodology to test the hypotheses. Section IV describes the IPO data and sample characteristics. Section V presents the empirical results. Section VI discusses the impact of IPO share retention in Taiwan on the choice between fixed price public offers and hybrid auctions, while Section VII concludes our study.

I. Related Literature on the Choice of IPO Underwriting Methods

Chemmanur and Liu (2006) analyze how an IPO firm selects an issue method between fixed price public offers and uniform price auctions when selling shares. Their model assumes that insiders have private information about the intrinsic firm value, and outsiders can produce information about the firm at a cost before bidding for shares in the IPO. Firm insiders seek to have a large number of outsiders producing information as this information is reflected in the secondary market price, providing a higher secondary market price for a higher intrinsic value firm.

In Chemmanur and Liu's (2006) model, a firm wishes to obtain higher proceeds from the sale of stock, but the firm also considers the secondary market price following the IPO. That is, the firm's goal is to maximize the dynamic objective function rather than minimize underpricing. When an IPO firm is young, small, faces a greater extent of information symmetry, or is selling only a small fraction of its equity in the IPO, the firm prefers using a fixed price public offer rather than an auction. This is because a fixed price public offer allows the IPO firm to induce the optimal extent of information production, increasing the importance of the secondary market price's impact.

Sherman (2005) models book-building, discriminatory auctions, and uniform price auctions in an environment where the number of investors and the accuracy of the investors' information are endogenous. (3) Her results indicate that book-building permits underwriters to control the allocation of the IPO shares allowing them to reduce the risk for both the issuers and the investors. It also allows them to control spending on information acquisition, thereby limiting either the underpricing or the aftermarket volatility. Her model suggests that an auction open to a large number of potential bidders is vulnerable to inaccurate pricing and to a fluctuation in the number of bidders.

Jagannathan et al. (2015) suggest that participating in auctions is substantially more difficult for investors when compared to a fixed price public offer or book-building. In addition, the complexity of auctions can lead to investor behavior that is undesirable for the issuer. For example, bidders must place their bids before they learn how many others will enter the sealed bid auction, which can be affected by return-chasing behavior (Chiang, Qian, and Sherman, 2010; Jagannathan et al., 2015).

The fluctuation in the number of participants greatly increases the variance of the auction outcome. In contrast, a fixed price public offer subscription level does not affect the offer price. The subscription level affects a person's chance of obtaining shares, but not his or her conditional return on obtaining those shares. Moreover, the bidders, who invest time and money in evaluating an IPO, risk being squeezed out by others who do not adequately understand the optimal bidding strategies, but place bids at very high prices. As such, they are effectively free-riding on informed investors' information.

Jagannathan et al. (2015) suggest that the optimal bid for any participant depends upon the number of other bidders, their information sets, and their bidding strategies. Since these three factors are unknown when the bids are placed in a sealed bid auction, this complexity presents a structural risk for both issuers and bidders. Jagannathan et al. (2015) cite the issue of this complexity as a reason for the apparent lack in popularity of IPO auctions around the world.

Ausubel (2002) suggests that IPOs should be sold through dynamic ascending-bid auctions, where the auctioneer names a price and the bidders respond with quantities. Dynamic auctions can reduce the winner's curse and avoid the inefficient outcomes generated by static sealed-bid auctions.

Biais and Faugeron-Crouzet (2002) present a theoretical model and use French IPOs to compare different IPO methods. They find that the French auction (the Mise en Vente) is superior to a fixed price public offer and the uniform price auction. (4) In addition, the Mise en Vente and the book-building method have similar incentive properties and can reach similar outcomes.

As for empirical studies, Jagannathan et al. (2015) investigate global trends for IPO methods and point out that IPO auctions have been tried in more than 25 of the 50 countries they surveyed. All of the countries, except India, Israel, the United States, and Vietnam, seem to have abandoned IPO auctions entirely.

Jagannathan et al. (2015) also analyze a sample of hybrid uniform-price IPO auctions in Singapore. In this case, they find evidence of return chasing and overbidding. They conclude that there is a large risk of failure for auctions due to the uncertainty about the number of bidders and the resulting adverse selection problem identified by Rock (1986). However, fixed price public offers may dominate auctions when it comes to maximizing revenue, inducing information gathering and facilitating transparency and the ease of implementation. When information gathering is a key factor, book-building is preferred, as it leads to better price discovery and lower underpricing.

Degeorge et al. (2010) examine 19 auctioned IPOs that used WR Hambrecht's Open IPO auction mechanism from 1999 to 2007. They find that bids submitted at high prices indicate that some investors (predominantly retail investors) try to free-ride on the mechanism. Furthermore, retail bids are less informative than institutional bids, but retail investors do not seem to affect the information extraction mechanism enough to discourage institutional investors from participating.

Demand from institutional investors is elastic, suggesting that institutional investors produce information and reveal it during the bidding process. (5) The pricing flexibility offered by this mechanism allows the issuer to buffer against free-riding by discounting the deal relative to the clearing price when some investors intentionally submit bids at high prices.

Kutsuna and Smith (2004) argue that the preference for book-building over the auction procedure in Japan is primarily because book-building and auctions are less costly for large and small issues, respectively. Book-building appears to have no aggregate size-weighted cost disadvantage relative to the Japanese hybrid auction method. At the same time, book-building has no opportunity cost of underinvestment and has the additional benefits associated with more accurate pricing.

Degeorge, Derrien, and Womack (2007) find that book-built IPOs were more likely to be followed and recommended by analysts associated with lead underwriters than IPOs using auctions. They suggest that the demise of IPO auctions in France can be explained by the concurrence of the difference in analyst coverage and an increase in the issuers' perceived importance of analyst coverage in the 1990s.

II. IPO Underwriting Process in Taiwan

Each IPO firm in Taiwan is required to file a registration statement with the Taiwan Securities and Futures Commission (TSFC) and must engage underwriters for due diligence before filing a preliminary prospectus. Once the TSFC approves the application, the issuer and the underwriters sign an underwriting contract. A tombstone announcement, which reports the final offer price, offer size, offering method, and summary of the issuer's financial information, is published in newspapers after the prospectus is registered and approved. (6)

IPO shares offered in Taiwan can be priced and allocated by three underwriting mechanisms: (1) a pure fixed price public offer, (2) a hybrid auction, and (3) a book-building offer. Both the hybrid auction and the book-building methods were allowed to be used beginning March 1995. Prior to the introduction of these two issue methods, the fixed price public offer was the only underwriting method. Book-building is allowed only for primary shares issued by an IPO firm, not for shares already owned by pre-IPO shareholders. Issuers felt that issuing primary shares in an IPO may lead to greater regulatory scrutiny and a lengthy review process. Therefore, issuers avoided administration procedures by selling the existing shares of the pre-IPO shareholders to the public. The proportion of pre-IPO shareholders' shares sold in an IPO to the number of shares owned by pre-IPO shareholders is usually between 10% and 20%.

An IPO firm could subsequently conduct a seasoned equity offering (SEO) within a few months following the IPO if the firm needs to raise new equity capital to finance its operation or expansion. As a result, book-building was rarely used except for Taiwan Depositary Receipts (TDRs) shares. Both fixed price public offers and hybrid auctions were dominant in the IPO market during our sample period.

Beginning in December 1999, regulators officially required underwriters to retain and purchase a proportion of IPO shares. (7) The purpose of this requirement was to reinforce underwriters' bargaining power and align underwriters' interests with those of IPO investors when the offer price is to be set. Underwriters are required to purchase between 10% and 25% of the IPO shares for regular IPOs and 50% for IPOs of special or risky high technology firms identified by the Ministry of Economic Affairs. The remaining shares are then sold to the public by either fixed price public offers or hybrid auctions.

There are several implications for requiring underwriters to purchase a portion of IPO shares. First, in addition to underwriter reputation, underwriter share retention also motivates underwriters to certify the offer price. It also forces them to take more underwriting risk if the offer price is set too high relative to market demand as underwriters are investors as well.

In addition, from the regulatory perspective of protecting individual retail investors, IPO share retention by underwriters tends to increase the underwriters' bargaining power in pricing IPOs with the IPO issuers due to the potential alignment of interests between underwriters and individual investors. Moreover, since underwriters have discretion when choosing the proportion of share retention between 10% and 25% for regular IPOs, the disclosed retention rate prior to the fixed price public offer sends a signal to individual investors when deciding whether to participate in an IPO subscription lottery (Chen, Jhou, and Yeh, 2007).

Finally, underwriter share retention allows underwriters to offset lower underwriting fees by purchasing underpriced IPO shares, earning the same return as the IPO investors (Chen et al., 2006). This gives underwriters a greater incentive to underprice IPO shares.

In a fixed price public offer, the lead underwriter and issuing firm jointly determines the offer price using a formula price suggested by the TSFC. The IPO firm can adjust the formula price upward or downward. (8) Each investor is usually allowed to subscribe to 1,000 shares (one round lot). A few exceptional cases allow the subscription of 2,000 shares. If an oversubscription occurs, shares are allocated using a public lottery employing computer-generated random numbers.

The TSFC implemented a new rule regarding the share subscription process to solve the problem of "token investors" in late 1997. The new rule states that the full payment amount for IPO shares must be deposited in investors' brokerage accounts when they submit the subscription orders for purchasing IPO shares. (9) This new rule also reduces the lottery processing fee allocated to underwriters from NTS30.0 to NTS 17.5 per application. In addition, investors who make multiple applications are disqualified from the IPO lottery draw. All investors have an equal opportunity to purchase IPO shares.

In the sequential hybrid auction, 50% of the IPO shares floated to the public are first sold by discriminatory auction in which investors pay what they bid if they win. The remaining 50% of the IPO shares are then sold to individual investors. In the auction tranche, the reservation price (minimum acceptable bid) is based on the formula price. However, the lead underwriter and the issuing firm have the right to negotiate the reservation price.

Investors who offer the highest bid gain allocation priority. Each qualified investor can submit one or more sealed bids for up to 6% of the auction tranche, which is 3% of the IPO shares offered in the public offering. Therefore, if institutional investors intend to acquire a large number of IPO shares, they are likely to participate in the auction tranche, rather than in the public offer tranche where one or two round lots are allowed for subscription.

For undersubscribed auctions, the offer price for a public subscription is fixed at the reservation price in the auction tranche. Any unsold shares in the auction tranche will be placed in the fixed price tranche. For oversubscribed auctions, if all winning bids are above 1.3 times the reservation price, the offer price in the public offer tranche is set at 1.3 times the reservation price. (10) However, if some shares are sold at winning bids less than 1.3 times the reservation price, then the offer price in the public offer tranche is set at the quantity-weighted average price of those winning bids no higher than 1.3 times the reservation price.

III. Hypothesis Development and Research Design

In this section, we develop two hypotheses to explore how issuers and underwriters choose between IPO underwriting methods available in Taiwan.

A. The Expected Net Wealth Gain Hypothesis

Based on the prospect theory advanced by Kahneman and Tversky (1979), Loughran and Ritter (2002) argue that pre-IPO shareholders feel complacent about leaving money on the table in IPOs by integrating the bad news, which is a relatively small loss from underpricing, with the good news, which is a relatively large gain in wealth. Further argument offers an alternative explanation for the existence of underpricing. That being said, it must be noted that prospect theory type preferences alone do not lead to underpricing. (11)

Inspired by the development of behavioral models in explaining IPO market behavior, we derive a measure of expected net wealth gain for pre-IPO shareholders in the spirit of the principle of wealth maximization. We hypothesize that this measure can explain the choice between IPO auctions and fixed price public offers. For ease of illustration, we call it the expected net wealth gain hypothesis.

In this study, the expected net wealth gain of pre-IPO shareholders is referred to as the wealth gain during the period between when the formula price is set and when the first closing price does not hit the price limit (hereafter the nonhit price). This is because a daily return limit of 7% in either direction is imposed on publicly traded stocks. We assume that pre-IPO shareholders collectively use the formula price as a reference point to measure their wealth for all held shares. Furthermore, in our sample, all shares offered in the IPOs are secondary shares, which are originally held and then sold by pre-IPO shareholders in public offerings.

Since the formula price is used as a reference point and all shares offered in the IPO are secondary shares, the expected net wealth gain of the pre-IPO shareholders has two components. The first component involves the number of shares retained by the pre-IPO shareholders, and is computed as the number of shares retained, multiplied by the change in the value per share from the formula price to the first nonhit closing price. The second component involves the number of shares sold in the IPO, and is computed as the number of shares sold by the pre-IPO shareholders in the IPO, multiplied by the change in the value per share from the formula price to the offer price. The expected net wealth gain is the sum of these two components.

We further normalize the net wealth gain by the expected wealth level of the pre-IPO shareholders prior to the IPO. In the case of a fixed price public offer, the normalized net wealth gain can be simplified as:

[NWG.sub.f,%] = [PD.sub.f] - ss x [UP.sub.f], (1)

where [PD.sub.f] = ([MP.sub.f] - FP)/FP, which is the normalized price difference for an IPO firm selecting the fixed price public offer. [MP.sub.f] is the first nonhit closing price for a fixed price public offer, FP is the formula price, and ss is the ratio of the number of secondary shares sold in the IPO to the number of all shares originally held by pre-IPO shareholders. [UP.sub.f] = ([MP.sub.f] - [O.sub.Pf])/FP, which is the normalized underpricing for an IPO firm selecting a fixed price public offer. [OP.sub.f] is the offer price for a fixed price public offer. The detail of derivation is relegated to Appendix A.

We then define the net wealth gain difference (NWGD) as the net wealth gain of the pre-IPO shareholders following the IPO if the IPO firm selects a fixed price public offer ([NWG.sub.f,%]) minus the net wealth gain of pre-IPO shareholders following the IPO if the IPO firm selects a hybrid auction ([NWG.sub.a,%]). More specifically, NWGD is the net increase in wealth if a fixed price public offer is used. Therefore, based on the principle of maximizing net wealth gain when choosing an issue method, we develop the following hypothesis:

Hypothesis 1: The expected net wealth gain hypothesis.

It is more likely for issuers to choose a fixed price public offer if their expected net wealth gain difference (NWGD) is positive.

B. The Expected Revenue Collection Hypothesis

We now examine the incentive of underwriters in the IPO underwriting process in Taiwan. If oversubscription occurs in a fixed price public offer, IPO shares are allocated by public lottery using computer-generated random numbers, and underwriters can earn the lottery processing fee from the IPO subscription. In contrast, if the hybrid auction procedure is selected, half of the IPO shares are first sold in the auction tranche. The use of a hybrid auction will result in a decrease in the available proportion of IPO shares in the fixed price tranche to 50% of the offering, compared to 100% of the offering for a pure fixed price public offer. Thus, for a given IPO, the revenues collected from the lottery processing fee for a hybrid auction offering would have been half as much as that collected from a fixed price public offer if the fixed price public offer had been selected.

According to Chen et al. (2006), underwriting fees account for a very small portion of underwriters' total revenue. Lottery processing fees and capital gains on retained IPO shares, in contrast, account for relatively large portions of their revenues. Taking the incentives of underwriters into account, we hypothesize that the difference in underwriters' revenues from lottery processing fees between the two procedures affects the choice of underwriting methods. This hypothesis is called the expected revenue collection hypothesis.

Similar to the expected net wealth gain, we define an underwriter's revenue difference (URD) as the underwriter's revenue if the IPO firm selects a fixed price public offer ([UR.sub.fi]) minus the underwriter's revenue if the IPO firm selects a hybrid auction ([UR.sub.ai]). Therefore, we develop the expected revenue collection hypothesis along with the monopoly power of the underwriters as follows:

Hypothesis 2: The expected revenue collection hypothesis

To increase the revenues collected from lottery processing fees, it is more likely that underwriters will persuade IPO issuers to choose a fixed price public offer if their revenue difference (URD) is positive.

C. Methodology

We assume that the choice of underwriting procedures is dependent upon the difference in expected net wealth gain, the difference in the underwriter's revenues, and other variables. The probit model is specified as follows:

[Y.sub.i] = [[beta].sub.0] + [[beta].sub.1] ([NWG.sub.fi] - [NWG.sub.ai]) + [[beta].sub.2] ([UR.sub.fi] - [UR.sub.ai]) + [gamma] (FC) + [delta] (Mkt) + [phi] (Control) + [e.sub.i], (2)

where [Y.sub.i] is the binary dependent variable and is equal to one when the hybrid auction procedure is selected, and zero otherwise. The variables, [NWG.sub.fi] and [NWG.sub.ai], are the expected net wealth gain (%) of the pre-IPO shareholders following the IPO if the IPO firm selects a fixed price public offer and a hybrid auction, respectively. Similarly, the variables [UR.sub.fi] and [UR.sub.ai] represent the respective underwriter's revenue (%) from the lottery processing fee for each offering type. FC represents a set of the firm's characteristics. Mkt is a set of variables related to the market conditions including the 30 trading day buy-and-hold market return prior to the first trading day and the standard deviation of the 30-trading day market return prior to the first trading day. (12) Other control variables include the year dummy variables.

The structural choice model (2) cannot be estimated directly as we cannot observe what the expected net wealth gain and underwriters' revenues would have been if the alternative underwriting procedure had been selected. In order to calculate the difference in the expected net wealth gain and the difference in the underwriters' revenues collected from the lottery processing fees between two underwriting procedures, we must estimate what the expected net wealth gain and underwriters' revenue would have been if the alternative underwriting procedure had been used. Since the issuing firm only selects one of the two underwriting procedures, we correct the self-selection bias using the two-stage estimation method proposed by Lee (1978).

From the derivation in Appendix A, we know that the expected net wealth gain difference (NWGD) normalized by the initial wealth can be stated as:

NWGD = ([PD.sub.f] - [PD.sub.a]) - ss([UP.sub.f] - [UP.sub.a]), (3)

where ss is the proportion of secondary shares sold in the IPO, and [PD.sub.f]([PD.sub.a]) and [UP.sub.f]([UP.sub.a]) is the normalized price difference and normalized underpricing for a fixed price public offer (hybrid auction), respectively. The normalized price difference is defined as [PD.sub.f] [approximately equal to] ([MP.sub.f] - FP)/FP for a fixed price public offer and [PD.sub.a] [approximately equal to] ([MP.sub.a] - FP)/FP for a hybrid auction, where [MP.sub.f] and [MP.sub.a] are the first nonhit closing prices for a fixed price public offer and for a hybrid auction offer, respectively, and FP is the formula price. The normalized underpricing is defined as [UP.sub.f] = ([MP.sub.f] - [OP.sub.f])/FP for a fixed price public offer and [UP.sub.a] = ([MP.sub.a] - ([OP.sub.a] + BP)/2)/FP for a hybrid auction, where [OP.sub.f] and [OP.sub.a] are offer prices for a fixed price public offer and for a hybrid auction offer, respectively, and BPis the quantity-weighted average winning bid price.

Therefore, we can calculate the difference in the expected net wealth gain ([NWG.sub.fi] - [NWG.sub.ai]) between the two underwriting procedures by estimating the respective normalized price difference, [PD.sub.f] and [PD.sub.a], as well as the respective normalized underpricing, [UP.sub.f] and [UP.sub.a]. Following Lee (1978), Dunbar (1995), and Chen, Dai, and Schatzberg (2010), we use the two-stage method of estimation to correct the self-selection bias.

The first stage in the two-stage method is to estimate the reduced-form probit model. The reduced-form is specified as follows:

[R.sub.i] = [[gamma].sub.0] + [[gamma].sub.1] [[PSI].sub.i] + [e.sup.*.sub.i], (4)

where [PSI] is the union of the variables related to the normalized price difference, normalized underpricing, and underwriter's revenues.

The predicted value [[??].sub.i] is used to calculate the inverse Mills ratio for adjusting the self-selection bias for the fixed price public offer and the hybrid auction:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

where f(x) and F(x) are the probability density function and cumulative distribution function of the standard normal distribution, respectively.

As for the second stage of the two-stage method, the regressions of the normalized price difference, normalized underpricing, and underwriter's revenues are augmented with the inverse Mills ratio as additional regressors. The estimate of the net wealth gain of the issuers is substituted into Model (2), which allows for the consistent estimation of the parameters by the maximum likelihood method.

Our expected net wealth gain hypothesis predicts that the coefficient on NWGD is significantly negative, since NWGD is the net increase in the wealth gain if the fixed price public offer procedure is used. Similarly, the expected revenue collection hypothesis predicts that the coefficient on URD is significantly negative, since URD is the net increase in the lottery processing fees if the fixed price public offer is selected.

IV. Data and Sample Characteristics

In this investigation, we analyze 712 Taiwanese IPOs. These IPOs are composed of firms initially listed on the Taiwan Stock Exchange (TWSE) or in Taiwan's over-the-counter market (OTC) from 1996 to 2003.

Of the 712 firms, 623 IPOs used the fixed price public offer. The remaining 89 IPO firms employed the hybrid auction procedure. In contrast to the sale of secondary shares at IPOs, primary share issuance was not common in Taiwan during our sample period as issuers felt that there was more regulatory scrutiny for issuing primary shares. There were only three hybrid auctions after the end of our sample period; one occurring in 2004, 2005, and 2008. (13)

Data in relation to the underwriting procedure, the lead underwriter, the offer price, the offer size, the subscription success rate, and the subscription level were collected from the Taiwan Securities Association. The underwriter share retention rate, formula price, and subscription date were collected from the tombstone announcements in newspapers. The auction data, including the number of shares sold in the auction tranche, the subscription level in the auction tranche, the reservation price, the maximum bidding price, the minimum bidding price, and the quantity-weighted winning price, were collected from the Taiwan Securities Association. The stock prices, market indices, and related financial data for the IPO firms were collected from the databank of the Taiwan Economic Journal (TEJ).

Figure 1 presents data concerning the annual distribution of IPO firms using either the fixed price public offers or hybrid auctions for the entire sample period. This data reveals that the use of hybrid auctions prevailed between 1997 and 1998. However, the fixed price public offer method is the predominant IPO underwriting procedure in Taiwan.

Table I presents the descriptive statistics for our IPO sample. The average and median amount of proceeds from the hybrid auction offerings are significantly larger than those of the fixed price public offers. (14) Since the amount of the proceeds and the number of subscriptions to IPO shares is positively related, the comparison of subscription levels between the two underwriting procedures is similar.

The average (median) underpricing is 27.4% (11.1%) for the fixed price public offerings compared to 16.3% (10.8%) for the hybrid auction offerings. The difference in the average underpricing between the two groups is significant, but the difference in the median underpricing is insignificant. The average values of the normalized price difference and of the normalized underpricing between the two underwriting methods are not significantly different. One exception is that the median of the normalized price difference for the fixed price public offer is significantly smaller than that of the hybrid auction.

The average earnings per share (EPS) and the book value per share for hybrid auction offerings are significantly larger than those for the fixed price public offers. As for firm age at the time of the IPO, the difference between both groups is not significant. The mean (median) proportion of the number of shares sold by pre-IPO shareholders in an IPO is 11% (10%) for the fixed price public offerings compared to 15% (15%) for the hybrid auction offerings. Since the auction offerings tend to be large offerings, their pre-IPO shareholders are required to sell more shares by regulation in Taiwan for the purpose of ownership diffusion and liquidity.

Adjustment is the percentage change from the formula price to the offer price for fixed price public offers. As such, the adjustment is used to account for the ability of the underwriter and issuer to predict the value of the IPO shares. It is not really the same as the partial adjustment of the offer price documented in Hanley (1993). (15) In the United States, the offer price revision is calculated from the midpoint of the initial offer price range reflecting the information of the underwriter and the issuer to the offer price. This offer price reflects the feedback from the investors during the marketing period.

In Taiwan, the adjustment is from a standard, mechanized formula price to the offer price. In this case, the offer price reflects the information of the underwriter and the issuer (i.e., equivalent to the midpoint of the initial offer price range in the United States).

The average 30-day buy-and-hold market return prior to the first trading day (MKM30) for fixed price public offerings is significantly larger than that for auctions, suggesting that IPO issuers tend to select the fixed price public offer when market conditions are good. Since the underwriting fee after 1999 is unavailable, we only report the results from 1996 to 1999. The average underwriting fee in Taiwan is 1.26%, and it is much lower than the 7% fee in the United States (Chen and Ritter, 2000). Furthermore, underwriters in Taiwan generally charge higher underwriting fees for hybrid auction offerings than for fixed price public offerings.

When compared to the underwriting fee, the lottery processing fee is the primary source of the underwriter's revenue. In the second half of our sample period, the lottery processing revenue decreased as the lottery processing fee per application had been reduced to NTS 17.5 from NTS30.0 since late 1997.

V. Empirical Results

In this section, we examine whether the difference in the issuer's expected net wealth gain and the difference in the underwriter's revenues collected from the lottery processing fees can explain who makes the choice in the underwriting procedure. We first report the projected difference in the issuer's expected net wealth gain and in the underwriter's revenues, and then examine their signs and significance in the structural choice model.

A. The Difference in the Expected Net Wealth Gain for Pre-IPO Shareholders

Since we cannot observe what the normalized price difference and the normalized underpricing would have been if the alternative underwriting procedure had been selected, we use the two-stage procedure proposed by Lee (1978) to address the self-selection problem and thereby correct the estimation bias. We regress the normalized price differences and normalized underpricing for the two groups of IPO firms on the characteristics of the issuing firms and the market conditions. We also consider the control variables of an underpricing regression that are used in Cliff and Denis (2004).

Table II reports the regression results. On average, electronics firms using a fixed price public offering yield an incremental normalized price difference of 0.2% as compared to 0.5% for the electronics firms using hybrid auctions. Moreover, the variable, prior market returns (MKM30), is significantly and positively related to the normalized price difference for the fixed price public offerings. This indicates that the price differences are likely to be correlated with the market conditions.

Prior market volatility (MKV30) is significantly and negatively related to the normalized price difference for both types of offerings. In addition, the variable, Adjustment, is significantly and positively related to the normalized price difference for both types of offerings. The significant inverse Mills ratio implies that the self-selection bias tends to exist for fixed price public offerings.

Furthermore, calculating the difference between the normalized price difference and normalized underpricing for firms using fixed price public offers, we find that this difference is exactly the variable, Adjustment. Since Adjustment is also an explanatory variable in the regression models, the coefficients on all of the variables except Adjustment are the same for the fixed price public offerings in Table II.

Although the property is not present for firms using the hybrid auction method, the signs and magnitude of the regression coefficients for the auction sample are similar. The coefficient on MKM30 suggests that the market change in the prior months is relevant to the aftermarket price. This result supports the findings of Derrien and Womack (2003) in that an IPO method is at a disadvantage when the offer price is set too far in advance.

Table III presents the results of the realized and projected normalized price difference, the realized and projected normalized underpricing, as well as the projected difference in the net wealth gain. We measure NWGD as the net increase in the pre-IPO shareholders' wealth gain when the fixed price public offer is used. The regression models in Table II are employed to obtain the projected values of the normalized price difference and normalized underpricing. Based on these projected values, in turn, we can calculate the projected difference in the net wealth gain for both groups of IPOs. In Panel B of Table III, the mean (median) projected difference of net wealth gain is around 0.1% (-31.6%) for fixed price public offering firms as compared to 117.1% (113.2%) for hybrid auction firms.

Furthermore, 83% of the fixed price public offering firms exhibit a positive NWGD, indicating that only 17% of the fixed price public offering firms would have been better off if they had chosen the hybrid auction option. In contrast, 76% of IPO firms using the hybrid auction demonstrate a positive NWGD, indicating that these IPO firms would have been better off if they had chosen the fixed price public offer.

Overall, we find that 83% of the fixed price public offering firms and 24% of the auctioned firms have already selected the most beneficial underwriting method available to them in terms of the expected net wealth gain difference. Among the 712 IPO firms, 587 firms (82%) were, or would have been, better off when choosing the fixed price public offering, while the remaining 125 firms (18%) were, or would have been, better off when choosing the hybrid auction option. (16)

B. The Difference in the Lottery Processing Revenues for Underwriters

We now examine the difference in the lottery processing revenues for underwriters if the alternative underwriting procedure is selected. Underwriters' lottery processing revenue is defined as the number of subscriptions to IPO shares in the fixed price public offer (or in the fixed price tranche of the hybrid auction), multiplied by the lottery processing fee per subscription. We estimate the level of the subscription to IPO shares for both the fixed price public offerings and the hybrid auction offerings using the characteristics of the issuing firms and the market conditions.

Table IV reports the regression results using the two-stage procedure developed by Lee (1978). The significantly positive coefficient on LnProceeds suggests that large offerings could attract more subscriptions from individual investors. The electronics firms are subject to market frenzies in our sample period and attract more subscriptions from individual investors. The prior market returns are significantly and positively related to the subscription level for both groups, indicating that investors are more willing to subscribe to IPO shares when market conditions are good. Moreover, the volatility of prior market returns is significantly and negatively related to the subscription level for both groups suggesting that investors are unwilling to subscribe to IPO shares in a volatile market.

The regulation change dummy (REG) is negatively related to the subscription level for both groups suggesting that the policy aimed at token investors was able to reduce the number of subscriptions and increase the likelihood of wining an IPO share allocation. The significantly positive coefficient on Adjustment for the fixed price public offers implies that investors view a positive price adjustment as a good signal. Both coefficients on the inverse Mills ratio are significantly positive indicating a significant self-selection bias for this sample.

In Table V, we compare the realized and projected subscription level in Panel A, as well as the realized and projected underwriters' revenues in Panel B. The models in Table IV are used to obtain the expected subscription level for IPO firms if these firms had adopted the alternative underwriting procedure. In Panel A, the average (median) realized subscription level is 594 (403) thousand round lots for the hybrid auction offerings, which is larger than the expected subscription level of 21 (7) thousand round lots had the issuing firms chosen the fixed price public offer method instead. The average (median) realized subscription level for the fixed price public offerings is 198 (114) thousand round lots, which is smaller than the expected subscription level of 436 (271) thousand round lots had the issuing firms elected to pursue a hybrid auction instead.

These results are not consistent with the notion that due to the different proportions of IPO shares available to individual investors in both underwriting procedures, the expected subscription level for IPOs using the hybrid auction would increase if they chose to pursue a fixed price public offer instead. In addition, the results differ from the idea that the expected subscription level for IPOs using the fixed price public offer would decrease if they chose to employ a hybrid auction instead.

In Panel B, we define the underwriters' revenues as the total subscription level multiplied by the lottery processing fees normalized by the amount of the proceeds. Since the amount of the proceeds raised from IPOs using a hybrid auction is usually larger than that of IPOs using a fixed price public offering, the average (median) realized revenue for IPOs using a fixed price public offering is 2.1% (0.9%). This is slightly higher than 1.9% (0.8%) for those IPOs using hybrid auctions.

For firms using a fixed price public offer, the average (median) projected underwriters' revenue would have been 4.8% (2.8%) if a hybrid auction had been used instead. For IPO firms using an auction, the average (median) projected underwriter's revenue would have been 0.04% (0.02%) if a fixed price public offer had been used. Thus, IPO firms using a hybrid auction would not have increased the subscription level or the underwriter's revenue if they had chosen the fixed price public offer method instead. In contrast, 78% of the fixed price public offering firms would have increased the subscription level and underwriter's revenue if the hybrid auction method had been selected. Overall, the preliminary evidence in Table V does not support the expected revenue collection hypothesis.

C. Determinants of Choosing an Underwriting Procedure

In Table VI, we use a probit model to explain the choice in underwriting procedures. The dependent variable is a binary variable with a value of one when a hybrid auction is used, and zero when a fixed price public offer is used. The control variables include the natural logarithm of proceeds (Sherman, 2005; Chemmanur and Liu, 2006) and the privatization dummy (Chemmanur and Liu, 2006). We also add the projected net wealth gain difference (PNWGD) between the fixed price public offer and the hybrid auction estimated in Table III and the projected underwriter's revenue difference (PURD) in Table V as explanatory variables.

Since the expected net wealth gain difference (NWGD) is the expected net increase in the wealth gain of pre-IPO shareholders if a fixed price public offering is used, the expected net wealth gain hypothesis predicts that the coefficient on the PNWGD is negative. Similarly, the underwriter's revenue difference (URD) is defined as the expected net increase in the underwriter's revenue if a fixed price public offer is selected. The expected revenue collection hypothesis also expects a significantly negative coefficient on PURD. Furthermore, we add year dummies to control for the time trend in the use of auctions.

Models (1) and (2) use the entire IPO sample to test the hypotheses, but they differ in the addition of the variable PURD. Model (3) has the same specification as Model (2), but restricts the sample to IPO firms listed on the TWSE. All of the models in Table VI report a significantly positive coefficient on LnProceeds. This evidence suggests that it is more likely for IPO firms with large offerings to select the hybrid auction method. These results are consistent with Chemmanur and Liu (2006) and Sherman (2005). (17)

The significantly negative coefficient on Privatization is inconsistent with the prediction made in Chemmanur and Liu (2006), who argue that IPO auctions may be optimal in the privatization of large state-owned firms. Further inspection of the data indicates that there are 11 privatization firms in our sample. Two of the privatization firms use the hybrid auction method during the most recent years after October 2000. Therefore, the recent privatization evidence seems to support their argument. Overall, it is still premature to test the hypothesis of Chemmanur and Liu (2006) using our privatization sample. Furthermore, the significantly negative coefficient on the projected net wealth gain difference (PNWGD) in all of the models supports the expected net wealth gain hypothesis in which IPO issuers choose the underwriting procedure that increases their expected wealth gains.

In Model (2) of Table VI, the coefficient on the projected underwriter's revenue difference (PURD) does not support the expected revenue collection hypothesis. We also rerun Model (2) for IPOs listed on the TWSE as a robustness check. The sample restriction greatly cuts back the number of fixed price public offers (137 fixed price public offerings vs. 49 auctions). Thus, the difference in the number of IPOs between the hybrid auction and fixed price public offerings would shrink substantially.

The results of Model (3) still support the expected net wealth gain hypothesis, but do not support the expected revenue collection hypothesis. Furthermore, we restrict our sample to the period from 1996 to 2000 given the fact that auctions nearly disappeared after the year 2000. The results (not reported) are robust and support the expected net wealth gain hypothesis, but not the expected revenue collection hypothesis.

Overall, we find that IPO issuers have more bargaining power in choosing an underwriting method than underwriters in Taiwan. The evidence regarding who chooses the offering method may play a role in understanding why IPO auctions have not been more popular around the world. Is it because underwriters have suppressed them or because issuers chose to not use them? We find that most IPO issuers did not choose to use auctions, which offers insight into the important issue of the IPO auction demise. However, this evidence cannot explain the total disappearance of auctions.

D. Characteristics of the Projected Sample of IPOs

Based on the structural probit Model (2) in Section III for the choice of underwriting procedure, we examine the characteristics of firms in the projected fixed price public offerings and the projected hybrid auction sample. In unreported results, the difference in the average IPO proceeds between the projected fixed price public offer and the projected auction is not found to be significant, but the median proceeds of the predictive auctioned IPOs are significantly larger than that of the predictive fixed price public offerings. This result is not consistent with the findings of Kutsuna and Simith (2004), who found that small firms in Japan prefer auctions to book-building from the perspective of the weighted average cost. Instead, we find that projected fixed price public offering firms exhibit larger underpricing than the projected auction firms.

The IPO firms in the group of projected fixed price public offers are associated with relatively high market volatility. This result is consistent with the risk aversion hypothesis, which predicts that higher firm risk and higher market volatility will lead the issuer to select the fixed price public offer method to raise a certain amount of proceeds (Benveniste and Busaba, 1997). Furthermore, underwriters would have earned the average lottery processing revenue of 2.2% from firms in the group of projected fixed price public offers and 1.3% from firms in the group of projected auction offers.

VI. Discussion

In this section, we discuss the possibility that underwriters encourage IPO issuers to choose the fixed price public offering to avoid the price risk faced by underwriters due to IPO share retention. Before IPO shares are sold to the public by either a fixed price public offer or a hybrid auction, underwriters are required to retain a certain percentage of IPO shares. More specifically, since December 1999, the retention requirement is 10% to 25% for ordinary IPO firms and 50% for young high-tech or special IPO firms. Underwriters pay the same offer price for the purchase of IPO retention shares as other investors do.

The offer price for IPOs using a fixed price public offer is predetermined prior to the offer day. In contrast, the offer price for IPOs using a hybrid auction is determined by the bidding result of the auction tranche. Thus, underwriters are facing the risk of an unexpected change in the offer price for IPOs using the hybrid auction if they have to purchase some portion of the IPO shares due to regulations. In other words, if the unexpected change in the offer price is significantly higher for IPOs using a hybrid auction, underwriters may not be willing to use a hybrid auction anymore because of the price risk concerns.

Furthermore, underwriters have an incentive to underprice IPO shares for the ease of marketing and capturing large capital gains simultaneously. Thus, underwriters prefer the fixed price public offer method to hybrid auctions for the ability to control the offer price through direct negotiation with IPO issuers. We hypothesize that underwriters encourage an IPO firm to choose fixed price public offerings to avoid the price risk faced by underwriters in purchasing IPO retention shares.

Since retaining IPO shares is a new institution to underwriters, they may apply prior experience to determine which procedure is more beneficial to them. Thus, we examine the IPO sample firms without IPO share retention. Panel A of Table VII reports that the average (median) price adjustment for hybrid auction offers is 55% (52%). This is significantly larger than the 1% (-2%) for fixed price public offers. Furthermore, 91% of IPOs using hybrid auctions exhibit a positive offer price adjustment indicating that the underwriters who underwrite these IPOs would pay higher offer prices than originally projected if they are required to buy IPO shares in advance. In contrast, only 45% of IPOs using the fixed price public offer method exhibit a pattern of positive offer price adjustment.

Although underwriters paid higher offer prices for IPOs using the hybrid auction, they were concerned with the aftermarket prices more than the offer prices. Since we do not know what underwriters expect the initial returns to be for IPOs using hybrid auctions ex ante, we examine the initial returns for both groups of IPOs ex post.

Panel A of Table VII indicates that the average (median) initial return for IPOs using fixed price public offers is 16% (10%) as compared to 26% (18%) for IPOs using hybrid auctions. (18) These results suggest that underwriters do not incur the loss immediately ex post. However, the average EPS also indicates that IPOs using hybrid auctions are firms with greater profitability than those using the fixed price public offer method. Overall, in the early period, when regulators did not require IPO share retention, the underwriters associated with IPOs using hybrid auctions did not face greater risks than those with IPOs using the fixed price public offer method.

Underwriter share retention allows underwriters to purchase underpriced IPO shares earning the same return as the IPO investors. This gives underwriters a greater incentive to underprice IPO shares. For the sample without the underwriter retention rate in Panel A, the average (median) price adjustment for the fixed price public offers is 1.1% (-2.0%). This is insignificantly different from the average (median) price adjustment of 32.7% (3.3%) for the fixed price public offers with the underwriter retention rate in Panel B.

The average initial return for fixed price public offers in Panel A is 16.3%. This is significantly smaller than the average return of 28.9% in Panel B. However, the median initial returns are not significantly different (10.0% in Panel A and 11.1% in Panel B). Therefore, for the fixed price public offer sample with the underwriter retention rate, there is some evidence that underwriters are willing to adjust offer prices higher for some IPOs and simultaneously earn larger initial returns. The evidence also suggests that underwriters in Taiwan are able to at least partially predict market demand and, as such, to appropriately adjust the offer price relative to the formula price.

In addition, if it is riskier for underwriters to use hybrid auctions, we expect that hybrid auction offers with underwriter retention will have a lower underwriter retention rate than their counterparts with fixed price public offers. In Panel B of Table VII, we compare the underwriter retention rates in both groups. In order to avoid IPOs with a regulated retention rate of 50%, confounding our empirical results, we only include IPOs with retention rates between 10% and 25% in the test. We find that the average (median) underwriter retention rate for IPOs using the hybrid auction method is 22.2% (25%), as compared to 20.4% (25%) for IPOs using fixed price public offerings. The insignificant difference in the underwriter retention rate between the two groups does not support the hypothesis that underwriters do not favor hybrid auctions (or favor the fixed price public offers) because of the high-risk underwriters face for their IPO share retention in the hybrid auction procedure. (19)

We find that the average (median) initial return for IPOs using the fixed price public offer is 28.9% (11.1%). This is significantly lower than 68.8% (43.6%) for IPOs using the hybrid auction method. The evidence is consistent with the results in Panel A of Table VII, which suggest that the risk facing underwriters using hybrid auctions is not relatively high. Therefore, it is less likely that underwriters encourage IPO firms to choose the fixed price public offer method to avoid the price risk associated with purchasing retained IPO shares in a hybrid auction.

VII. Conclusion

In this paper, we propose two hypotheses, the expected net wealth gain hypothesis and the expected revenue collection hypothesis, to explain the choice between the use of fixed price public offerings and hybrid auctions for IPOs in Taiwan from 1996 to 2003. To measure shareholder motivation, we develop a measure of the difference in the expected net wealth gain for pre-IPO shareholders between fixed price public offerings and auctions. We argue that the IPO issuers choose the offering procedure that provides them with a larger expected net wealth gain.

We find that 83% of IPO issuers who chose the fixed price public offer method would have been worse off had they selected the hybrid auction method instead. The results imply that most IPO issuers using fixed price public offerings chose the optimal issue method given their firm characteristics. In contrast, 76% of IPO issuers who chose the hybrid auction method would have been better off had they selected the fixed price public offer, suggesting that most IPO issuers using hybrid auctions should instead choose fixed price public offerings, given their firm characteristics. Consistent with the principle of wealth maximization, most IPO issuers in Taiwan chose the fixed price public offer method based on their expected net wealth gains, relative to their reference points, which are the formula prices set prior to their IPOs.

Furthermore, an unusual feature of Taiwan's IPOs is the difference in fees. The subscription fees or lottery processing fees first appear to be biased in favor of fixed price public offerings as the fees are collected for only half of the shares in hybrid auctions. Given that auctions tended to have substantially higher subscription levels, total underwriter revenues actually favor auctions. However, auctions have not been chosen very often implying that underwriters are not able to control the choice of the issue method.

Overall, our study suggests that the perspective of expected wealth gain has some explanatory power when explaining IPO market behavior. Rather than underwriters directing the choice based on underwriter revenues, our study suggests that IPO issuers appear to be dominant when choosing between the hybrid auction and the fixed price public offer methods. The evidence regarding who chooses the offering method may play a role in determining why IPO auctions have not been more popular around the world. Is it because underwriters have suppressed them or because issuers chose to not use them? Our study suggests that most IPO issuers elect to avoid the use of auctions. Our results may also provide some insights into the IPO auction demise phenomenon, although we note that our evidence cannot completely explain the total disappearance of auctions in many countries.

Appendix A: The Expected Net Wealth Gain for Pre-IPO Shareholders

We define the expected net wealth gain of pre-IPO shareholders as the wealth gain experienced during the period between when the formula price is set and when the first closing price does not hit the price limit. Note that the formula price serves as a reference price of the stock held by pre-IPO shareholders.

In our sample, all shares offered in the IPOs are secondary shares originally held and then sold by pre-IPO shareholders in public offerings. Thus, we compute the expected net wealth gain as the number of shares retained by pre-IPO shareholders multiplied by the change in the value per share from the formula price to the first closing price that does not hit the price limit, plus the shares sold by pre-IPO shareholders in the IPO multiplied by the change in the value per share from the formula price to the offer price. [NWG.sub.f] ([NWG.sub.a]) is the expected net wealth gain of pre-IPO shareholders following the IPO if an IPO firm selects the fixed price public offer (hybrid auction) method. We hypothesize that the pre-IPO shareholders of an IPO firm select the firm's underwriting procedure based on the expected net wealth gain when making the decision. Note that this hypothesis for decision making based on the expected net wealth gain is consistent with the prospect theory developed by Kahneman and Tversky (1979) who find that individuals make choices under uncertainty by maximizing a value function that assesses wealth changes. The definition of all of the variables is given as follows:

SS is the number of secondary shares sold by pre-IPO shareholders in the IPO.

TS is the total number of shares held by pre-IPO shareholders.

FP is the formula price of the IPO stock. Thus, the expected wealth of pre-IPO shareholders prior to the IPO is [W.sub.0] = (TS) - (FP).

[OP.sub.f] ([OP.sub.a]) is the offer price if the IPO firm selects the fixed price public offer (hybrid auction) method.

[MP.sub.f] ([MP.sub.a]) is the first closing price that does not hit the price limit following the IPO if the IPO firm selects the fixed price public offer (hybrid auction) method.

BP is the quantity-weighted average winning bid price in the hybrid auction.

We compute the expected net wealth gain for the IPO firm selecting the fixed price public offer method as:

[NWG.sub.f] = (TS - SS) x ([MP.sub.f] - FP) + SS x ([OP.sub.f] - FP), = (TS - SS) x [MP.sub.f] + SS x [OP.sub.f] - TS x FP, = TS ([MP.sub.f] - FP) - SS ([MP.sub.f] - [OP.sub.f]).

We further normalize [NWG.sub.f] using the expected wealth of pre-IPO shareholders prior to the IPO and obtain:

[NWG.sub.f,%] [NWG.sub.f]/(TS x FP) = ([MP.sub.f] - FP/FP) - ss x ([MP.sub.f] - [OP.sub.f]/FP) = [PD.sub.f] - ss x [UP.sub.f],

where [PD.sub.f] = ([MP.sub.f] - FP)/FP is the normalized price difference for an IPO firm selecting the fixed price public offer method, ss = SS/TS is the ratio of the number of secondary shares sold in the IPO to the number of all shares originally held by pre-IPO shareholders, and [UP.sub.f] = ([MP.sub.f] - [OP.sub.f])/FP is the normalized underpricing for an IPO firm selecting the fixed price public offer method.

Similarly, we calculate the expected net wealth gain for an IPO firm selecting the hybrid auction method as:

[NWG.sub.a] = (TS - SS) x [MP.sub.a] + SS x ([OP.sub.a] + BP)/2 -TS x FP = TS x ([MP.sub.a] - FP) -SSx [[MP.sub.a] - ([OP.sub.a] + BP)/2].

We also normalize [NWG.sub.a] using the expected wealth of pre-IPO shareholders prior to the IPO and obtain:

[NWG.sub.a,%] = [NWG.sub.a]/(TS x FP) = ([MP.sub.a] - FP/FP) - ss x [[MP.sub.a] - ([OP.sub.a] + BP)/2/FP] = [PD.sub.a] - ss x [UP.sub.a],

where [PD.sub.a] = ([MP.sub.a] - FP)/FP is the normalized price difference for an IPO firm selecting the hybrid auction method, and [UP.sub.a] = ([MP.sub.a] - ([OP.sub.a] + BP)/2)/FP is the normalized underpricing for an IPO firm selecting the hybrid auction.

Finally, to test the hypothesis that the pre-IPO shareholders of an IPO firm select the IPO firm's underwriting procedure based on the expected net wealth gain, we calculate the difference in normalized net wealth gain between the two offering procedures and obtain:

NWGD = [NWG.sub.f,%] - [NWG.sub.a,%] = ([PD.sub.f] - [PD.sub.a]) - ss([UP.sub.f] - [UP.sub.a]).

NWGD is an increase in the expected net wealth gain of pre-IPO shareholders if the fixed price public offer is selected. According to the hypothesis, if NWGD is greater than zero, the pre-IPO shareholders of an IPO firm will select the fixed price public offer instead of the hybrid auction.

Appendix B: Variable Descriptions

Variable                                Description

Proceeds              For a fixed price public offer, it is defined
                        as the number of IPO shares multiplied by
                        the offer price and measured in millions of
                        NT$. The exchange rate is approximately
                        NT$34/$ at the end of 2003. For a hybrid
                        auction, it is defined as the number of
                        shares in the auction tranche multiplied by
                        the quantity-weighted winning price plus the
                        number of shares in the fixed price tranche
                        multiplied by the offer price. LnProceeds is
                        the natural logarithm of the proceeds in
                        millions of NTS.
Subscription level    The number of subscriptions for the fixed price
                        public offer or the fixed price tranche in
                        the hybrid auction.
Underpricing          For IPOs using the fixed price public offer,
                        underpricing is the percentage return
                        calculated from the offer price to the first
                        closing price that does not hit the daily
                        price limit. For IPOs using the hybrid
                        auction, underpricing is the percentage
                        return calculated from the midpoint of the
                        quantity-weighted winning price and offer
                        price to the first non-hit closing price.
Normalized price      Based on Appendix A, the normalized price
  difference            difference is defined as [PD.sub.f] =
                        ([MP.sub.f]--FP)/FP for a fixed price public
                        offer and [PD.sub.a] = ([MP.sub.a]--FP)/FP
                        for a hybrid auction offer, where [MP.sub.f]
                        and [MP.sub.a] are the first nonhit closing
                        prices for a fixed price public offer and for
                        a hybrid auction offer, respectively. FP is
                        the formula price.
Normalized            The normalized underpricing is defined as
  underpricing          [UP.sub.f] = ([MP.sub.f]--[OP.sub.f])/FP for
                        a fixed price public offer and [UP.sub.a] =
                        ([MP.sub.a]--(BP + [OP.sub.a])/2)/FP for a
                        hybrid auction offer, where [0P.sub.f] and
                        [OP.sub.a] are the offer prices for a fixed
                        price public offer and for a hybrid auction
                        offer, respectively. BP is the
                        quantity-weighted average winning bid price.
EPS                   The annual average of earnings per share for
                        three years before the IPO date.
Book Value            The annual average of the book equity value per
                        share for three years prior to the IPO date.
Age                   The firm's age in the IPO year.
Shares sold           The ratio of the number of secondary shares
                        sold in the IPO to the number of all shares
                        originally held by pre-IPO shareholders.
Adjustment            The percentage change from the formula price to
                        the offer price for fixed price public offers,
                        and from the formula price to the reservation
                        price for hybrid auction offers.
MKM20                 The buy-and-hold market index return over the
                        20 trading days before the subscription date.
MKM30                 The buy-and-hold market index return over the
                        30 trading days before the first IPO trading
                        day.
MKV20                 The daily standard deviation of market index
                        returns over the 20 trading days before the
                        subscription date.
MKV30                 The daily standard deviation of market index
                        returns over the 30 trading days before the
                        first IPO trading day.
Underwriting fee      Underwriting fee is expressed as a percentage
                        of proceeds.
Lottery processing    Lottery processing revenue is computed as the
                        total subscription level of a fixed price
  revenue               public offer tranche multiplied by the
                        lottery processing fee per application. We
                        normalize the underwriters' revenue by the
                        amount of proceeds. The lottery processing
                        revenue is calculated as follows due to the
                        regulation change: (a) between April 1, 1995
                        and November 11, 1997, the number of
                        subscriptions multiplied by NT$30.0; (b)
                        after November 11, 1997, the number of
                        subscriptions multiplied by NT$17.5.
TWSE                  A dummy variable set equal to one if the firm
                        is listed on the Taiwan Stock Exchange, and
                        zero otherwise.
Electronics           Electronics is the proportion of the IPO firms
                        in the electronics industry.
ELE                   A dummy variable set equal to one if the firm
                        is in the electronics industry, and zero
                        otherwise.
Privatization         A dummy variable set equal to one when the firm
                        is a privatization firm, and zero otherwise.
REG                   A regulation dummy variable set equal to one
                        for IPOs after November 1997, and zero
                        otherwise.
PNWGD                 The projected net wealth gain difference
                        between the fixed price public offer and the
                        hybrid auction estimated in Table III.
PURD                  The projected underwriters' revenue difference
                        in the lottery processing fee between the
                        fixed price public offer and hybrid auction
                        estimated in Table V
Offer price           The percentage change from the formula price to
  adjustment            the offer price.
Traditional initial   The percentage change from the offer price to
  return                the first closing price that does not hit the
                        daily price limit.
Underwriter           The percentage of IPO shares retained and
  retention rate        purchased by underwriters. Underwriters in
                        Taiwan have been required to retain 10% to
                        25% of the IPO shares for their own
                        purchases, and to retain 50% for the
                        offerings of special or risky high technology
                        firms since December 1999.


References

Ausubel, L., 2002, "Implications of Auction Theory for New Issues Markets," Brookings-Wharton Papers on Financial Services 5, 313-343.

Benveniste, L. and W. Busaba, 1997, "Bookbuilding vs. Fixed Price: An Analysis of Competing Strategies for Marketing IPOs," Journal of Financial and Quantitative Analysis 32, 383-403.

Benveniste, L. and P. Spindt, 1989, "How Investment Bankers Determine the Offer Price and Allocation of New Issues," Journal of Financial Economics 24, 343-361.

Biais, B. and A. Faugeron-Crouzet, 2002, "IPO Auctions: English, Dutch, ... French, and Internet," Journal of Financial Intermediation 11, 9-36.

Bower, N., 1989, "Firm Value and the Choices of Offering Method in Initial Public Offerings," Journal of Finance 44, 647-662.

Bradley, D. and B. Jordan, 2002, "Partial Adjustment to Public Information and IPO Underpricing," Journal of Financial and Quantitative Analysis 37, 595-616.

Chemmanur, T. and M. Liu, 2006, "How Should A Firm Go Public? A Dynamic Model of the Choice Between Fixed-Price Offerings and Auctions in IPOs and Privatizations," Carroll School of Management, Boston College Working Paper.

Chen, H-C., S-S. Chen, and C-W. Huang, 2012, "Why Do Insiders Sell Shares Following IPO Lockups?" Financial Management 41, 813-847.

Chen, H-C., N. Dai, and J. Schatzberg, 2010, "The Choice of Equity Selling Mechanisms: PIPEs versus SEOs," Journal of Corporate Finance 16, 104-119.

Chen, H-C., R. Fok, and Y-J. Wang, 2006, "Why Do Underwriters Charge Low Underwriting Fees for Initial Public Offerings in Taiwan?" Journal of Business Finance and Accounting 33, 979-1005.

Chen, H-C., C-J. Jhou, and H-C. Yeh, 2007, "Signaling by Underwriter Retention Rate in the IPO Market," Applied Economics 39, 1973-1983.

Chen, H-C. and J. Ritter, 2000, "The Seven Percent Solution," Journal of Finance 55, 1105-1131.

Chiang, Y-M., Y. Qian, and A. Sherman, 2010, "Endogenous Entry and Partial Adjustment in IPO Auctions: Are Institutional Investors Better Informed?" Review of Financial Studies 23, 1200-1230.

Chowdhry, B. and A. Sherman, 1996, "International Difference in Oversubscription and Underpricing of IPOs," Journal of Corporate Finance 2, 359-381.

Cliff, M. and D. Denis, 2004, "Do Initial Public Offering Firms Purchase Analyst Coverage with Underpricing?" Journal of Finance 59, 2871-2901.

Degeorge, F., F. Derrien, and K. Womack, 2007, "Analyst Hype in IPOs: Explaining the Popularity of Bookbuilding," Review of Financial Studies 20, 1021-1058.

Degeorge, F., F. Derrien, and K. Womack, 2010, "Auctioned IPOs: The US Evidence," Journal of Financial Economics 98, 177-194.

Derrien, F. and K. Womack, 2003, "Auctions vs. Bookbuilding and the Control of Underpricing in Hot IPO Markets," Review of Financial Studies 16, 31-61.

Dunbar, C., 1995, "The Use of Warrants as Underwriter Compensation in Initial Public Offerings," Journal of Financial Economics 38, 59-78.

Dunbar, C., 1998, "The Choice Between Firm-Commitment and Best-Effort Offering Methods in IPOs: The Effect of Unsuccessful Offers," Journal of Financial Intermediation 7, 60-90.

Ghosh, C., M. Petrova, Z. Feng, and M. Pattanapanchai, 2012, "Does IPO Pricing Reflect Public Information? New Insights from Equity Carve-Outs," Financial Management 41, 1-33.

Hanley, K., 1993, "Underpricing of Initial Public Offerings and the Partial Adjustment Phenomenon," Journal of Financial Economics 37, 239-257.

Hsu, Y. and C-Y. Shiu, 2010, "The Overconfidence of Investors in the Primary Market," Pacific-Basin Finance Journal 18, 217-239.

Jagannathan, R., A. Jirnyi, and A. Sherman, 2015, "Share Auctions of Initial Public Offerings: Global Evidence," Journal of Financial Intermediation, forthcoming.

Kahneman, D. and A. Tversky, 1979, "Prospect Theory: An Analysis of Decision under Risk," Econometrica 47, 263-291.

Kutsuna, K. and R. Smith, 2004, "Why Does Book Building Drive Out Auction Methods of IPO Issuance? Evidence from Japan," Review of Financial Studies 17, 1129-1166.

Lee, L-F., 1978, "Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables," International Economic Review 19, 415-433.

Liu, L., A. Sherman, and Y. Zhang, 2014, "The Long-Run Role of the Media: Evidence from Initial Public Offerings," Management Science 60, 1945-1964.

Liu, Y-J., K-C. Wei, and G. Liaw, 2001, "On the Demand Elasticity of Initial Public Offerings: An Analysis of Discriminatory Auctions," International Review of Finance 2, 151-178.

Loughran, T. and J. Ritter, 2002, "Why Don't Issuers Get Upset About Leaving Money on the Table in IPOs?" Review of Financial Studies 15, 413-443.

Lowry, M. and G.W. Schwert, 2004, "Is the IPO Pricing Process Efficient?" Journal of Financial Economics 71, 3-26.

Mandelker, G. and A. Raviv, 1977, "Investment Banking: An Economic Analysis of Optimal Underwriting Contracts," Journal of Finance 32, 683-694.

Pagano, M., 1993, "The Flotation of Companies on the Stock Market: A Coordination Failure Model," European Economic Review 37, 1101-1125.

Pagano, M., F. Panetta, and L. Zingales, 1998, "Why Do Companies Go Public? An Empirical Analysis," Journal of Finance 53, 27-64.

Ritter, J., 1984, "The "Hot Issue" Market of 1980f Journal of Business 57, 215-240.

Rock, K., 1986, "Why New Issues Are Underpriced?" Journal of Financial Economics 15, 187-212.

Sherman, A., 2005, "Global Trends in IPO Methods: Book Building versus Auctions with Endogenous Entry," Journal of Financial Economics 78, 615-649.

Sherman, A. and S. Titman, 2002, "Building the IPO Order Book: Underpricing and Participation Limits with Costly Information," Journal of Financial Economics 65, 3-29.

Hsuan-Chi Chen and Sheng-Ching Wu *

* Hsuan-Chi Chen is an Associate Professor in the Anderson School of Management at the University of New Mexico in Albuquerque, NM. Sheng-Ching Wu is an Assistant Professor in the Department of Finance at Da-Yeh University in Changhua, Taiwan.

We are grateful to Bill Christie (former Editor), Raghu Rau (Editor), and an anonymous referee for their insight and guidance. We also thank Chaoshin Chiao, Chung-Chien Wang, Tjalling van der Goot, Te-Chung Hu, Jordan Levy, Jay Ritter, Alexandria Solt, Hsiu-Chuan Yeh, and the seminar participants at the National Taiwan University International Conference on Finance, the Annual Conference on the Theories and Practices of Securities and Financial Markets, Shih Chien University, and Tamkang University for the helpful comments. An earlier draft of the paper received a best paper award granted by the Securities & Futures Institute in Taiwan. Hsuan-Chi Chen gratefully acknowledges support from the Anderson School of Management at the University of New Mexico. The authors also gratefully acknowledge the editing assistance of Wendy Jennings.

(1) Among 23 auction IPOs from 1999 to 2013 in the United States, WR Hambrecht was the lead underwriter in 20 of them. Three auction IPOs were lead managed by traditional investment banks. Google was lead managed by both Credit Suisse First Boston and Morgan Stanley in 2004. NetSuite was lead managed by Credit Suisse in 2007. Rackspace Hosting was lead managed by Credit Suisse, Goldman Sachs, and Merrill Lynch in 2008. This suggests that traditional investment banks have not completely refused to lead auction IPOs.

(2) The income structure for underwriting IPOs in Taiwan is relatively unique, as the underwriting fees, the IPO lottery processing fees, and the capital gains from underwriter share retention are the three main sources of revenue to underwriters. However, data for the IPO underwriting fees after 1999 are not publicly available. Lottery processing fees declined after the regulatory change in 1997. Thus, lottery processing fees have become less important as a source of income to underwriters, but they still exceed underwriting fees (Chen et al., 2006). Due to the limited data availability of underwriting fees in our IPO sample, we estimate the lottery processing fees as a main source of underwriter revenues. We further discuss IPO share retention and its impact on the choice of issue methods in Sections II and VI.

(3) In discriminatory auction offerings, all winning bidders pay what they bid. In uniform price auction offerings, all winning bidders pay the same offer price. The common offer price may be the market-clearing price (the highest price at which all shares are sold), or it may be below the clearing price, leading to "leave something on the table" (Ausubel, 2002; Sherman, 2005; Jagannathan et al., 2015).

(4) In the French auction (Mise en Vente) offerings, investors submit limit orders as in a sealed-bid auction. The auctioneer computes a cumulative demand curve after collecting the bids. The issuer and the underwriter then negotiate with the auctioneer to choose a common offer price that every selected investor will pay for his shares. Investors receive shares on a pro rata basis. Derrien and Womack (2003) offer more detail about the French auction.

(5) The results on bid elasticity are difficult to interpret as WR Hambrecht has reserved the right to communicate "general auction trends" to institutional, but not individual investors in many of their auctions. In other words, the underwriter reserves the right to contact major institutional investors and advise them to either increase or decrease their bids, based on the bidding trends of others. Thus, by examining only the final bids in those auctions, after those communications had occurred, it is difficult to determine whether the main information flow was from institutional investors to the underwriter or from the underwriter to the institutional investors. We thank the referee for pointing this out.

(6) IPOs in Taiwan have been underwritten primarily by either the best efforts or the standby commitment method. Since 1995, when government policy shifted and some issuers failed to raise equity capital using the best efforts method, the standby commitment method has been dominant in the IPO market. Standby agreements are, for all practical purposes, firm commitments. They are technically different, but both are ways in which the proceeds are guaranteed by the underwriter, as opposed to a best efforts contract in which there is no guarantee. Standby agreements are the more common approach internationally (outside of the United States).

(7) Before the regulation was officially imposed, some underwriters had been able to purchase some portion of the IPO shares through negotiations with IPO issuers. Underwriters also disclosed the proportion and number of shares in the tombstone announcements published in newspapers.

(8) The TSFC pricing formula is as follows: P = EPS x (PIE) x (0.4) + DIV x (0.2) + BV x (0.2) + EDIV x (0.2), where P = formula price, EPS = the most recent three-year average earnings per share, P/E = the most recent three-year average price-earnings ratio of three comparable public firms in a similar industry, DIV = the most recent three-year average dividend per share of the issuing firm divided by the average of the most recent three-year dividend yields of three comparable public firms in a similar industry, BV = IPO issuer's book value per share, and EDIV = IPO issuer's one-year projected dividend divided by the one-year prime interest rate.

(9) The rules to limit token investors are similar to the payment in advance rules that are common for fixed price public offers in most countries. Chowdhry and Sherman (1996) argue that if the issuer collects the interest float on checks deposited by investors for shares they bid, this interest revenue reduces the cost of underpricing and, as such, increases the incentive to underprice the offering further. However, the full payment amount for IPO shares in Taiwan must be deposited in investors' brokerage accounts, rather than some designated account. Thus, Taiwanese rules do not benefit either issuers or underwriters, which is in contrast to the rules in other countries.

(10) Essentially, the reservation price in the auction tranche and 1.3 times the reservation price serve as the minimum and maximum offer prices for the fixed price tranche. Prior to September 30, 1999, the multiple used for determining the offer price in the fixed price tranche was 1.5. See Liu, Wei, and Liaw (2001) and Chen et al. (2006) for more detail.

(11) Loughran and Ritter (2002) also assume that underwriters are able to collude with some investors to receive kickbacks on the allocation of the underpriced shares. Thus, prospect theory preferences only lead to underpricing when the parameters are such that the kickbacks from some investors are greater than the increased fee (as a percentage of the proceeds) from setting a higher offer price.

(12) The first trading day is referred to as the IPO exchange-listing day when IPO shares begin trading. The 30-day trading period prior to the first trading day always overlaps with the offer period (Hsu and Shiu, 2010). Because of the daily price limit of 7% in Taiwan, we define underpricing for IPOs using the fixed price public offer as the percentage difference computed from the offer price to the first nonhit closing price. Note that the first nothit day is not necessarily the first trading day or the IPO exchange-listing day.

(13) These three hybrid auction offerings are Taiwan Cooperative Bank in 2004, PChome Online in 2005, and CSBC Corp., a government-owned shipbuilder, in 2008. Furthermore, the book-building procedure has become increasingly popular since 2004 because a regulation change in Taiwan removed the restriction of issuing primary shares in book-built offerings (Chiang et al., 2010).

(14) The exchange rate was approximately NT$34/US$ at the end of 2003. Also, for brevity, Appendix B provides the definition of each variable. In Table I, if we exclude the privatization IPO, Chunghwa Telecom, from the hybrid auction offerings because of its enormous proceeds of NTS 153 billion in 2000, the average amount of the proceeds would decrease from NTS 2,854 million to NTS 1,146 million. However, the empirical results are not materially changed if we exclude this IPO in a subsequent analysis.

(15) One explanation for the partial adjustment of the offer price is that underwriters reward investors who truthfully reveal their private information in the book-building period by partially incorporating the information into the offer price (Benveniste and Spindt, 1989; Hanley, 1993). In contrast, however, Loughran and Ritter (2002), Bradley and Jordan (2002), Lowry and Schwert (2004), and Ghosh et al. (2012) present evidence that the offer price update partially adjusts to both private and public information. This is consistent with the prediction provided in Loughran and Ritter (2002). These results are also consistent with the costly information production model in Sherman and Titman (2002). In the Benveniste and Spindt (1989) model, information is endowed and the underwriter needs only to induce the investors to reveal that information. However, if there is an opportunity cost for acquiring the information first, and that cost is also related to recent stock market returns, the costly information production model then explains the partial adjustment to the public information as well.

(16) Note that there were only three hybrid auctions in Taiwan after 2003. This information is consistent with Jagannathan et al. (2015) in that all of the countries except India, Israel, the United States, and Vietnam seem to have abandoned IPO auctions entirely. The book-building procedure has become increasingly popular since 2004 as the regulation change removed the restriction of issuing primary shares for book-built offerings (Chiang et al., 2010).

(17) Sherman (2005) finds that book-building can be used to induce a greater level of scrutiny than auctions. Chemmanur and Liu (2006) illustrate that fixed price public offers can induce more evaluation about an offering than auctions. Thus, both models predict that auctions may not be able to guarantee a sufficient level of scrutiny, especially for small, young IPO firms.

(18) Underpricing for the entire sample of 89 auctions has an average of 16.31% (Table I), but traditional initial returns in the two subsamples (without the underwriter retention rate in the sample period and with the underwriter retention rate in 2000 to 2003) in Table VII have averages of 25.72% (Panel A) and 68.76% (Panel B). Note that the definitions for these two variables are different between Tables I and VII. For IPOs using the hybrid auction method, the traditional initial return is the percentage difference from the offer price to the first nonhit closing price. However, underpricing in Table I is defined as the percentage difference from the midpoint of the quantity-weighted winning price and the offer price to the first nonhit closing price. For the sample of the fixed price public offers, the definitions for the underpricing and traditional initial return are the same.

(19) Alternatively, we obtain evidence as to which issue method underwriters believe to contain greater underpricing by examining their retention rates. All else being equal, underwriters are more willing to retain shares in an IPO that they expect to be more heavily underpriced. The lack of a significant difference in underwriter retention rates is consistent with underwriters' limited expectations regarding any one method leading to a substantially higher underpricing. We thank the referee for providing this argument.

Table 1. Descriptive Statistics

This table provides descriptive statistics for the sample of 712
Taiwanese IPOs from January 1996 to December 2003. All of the
variables are defined in Appendix B. The median value of each
variable is reported in brackets. The standard deviation of each
variable is reported in parentheses

                                                    Fixed
                                       All       Price Offer

                                      Mean          Mean
                                    [Median]      [Median]
                                   (Std. Dev.)   (Std. Dev.)

Proceeds (millions of NTS)            676.45       365.37
                                     [195.04]     [168.53]
                                    (5794.31)     (788.23)
Subscription level (thousand)         247.48       197.99
                                     [131.40]     [113.96]
                                     (373.12)     (294.52)
Underpricing (%)                       25.98        27.37
                                      [11.11]      [11.11]
                                      (49.17)      (51.02)
Normalized price difference (%)        75.49        73.27
                                      [28.84]      [21.12]
                                     (211.83)     (224.05)
Normalized underpricing (%)            39.49        40.95
                                      [12.33]      [11.04]
                                     (110.48)     (115.79)
EPS                                     2.06         2.01
                                       [1.78]       [1.75]
                                       (1.53)       (1.49)
Book Value                             14.59        14.48
                                      [13.56]      [13.48]
                                       (4.06)       (3.91)
Age (year)                             15.60        15.53
                                      [13.00]      [13.00]
                                       (9.58)       (9.73)
Shares sold (%)                        11.64        11.12
                                      [10.00]      [10.00]
                                       (5.38)       (4.85)
Adjustment (%)                         28.69        32.32
                                       [1.30]       [3.01]
                                     (141.72)     (150.90)
MKM30 (%)                               0.96         1.23
                                       [0.65]       [0.75]
                                      (10.41)      (10.36)
MKV30 (%)                               1.61         1.63
                                       [1.61]       [1.63]
                                       (0.51)       (0.51)
Underwriting fee (%), 1996 to           1.26         1.19
  1999                                 [1.01]       [1.00]
                                       (0.73)       (0.71)
Lottery processing revenue              3.00         3.30
  (%), 1996 to 1999                    [1.10]       [1.16]
                                       (5.51)       (6.11)
Lottery processing revenue              1.48         1.50
  (%), 2000 to 2003                    [0.77]       [0.77]
                                       (1.94)       (1.97)
TWSE listing (%)                       26.12        21.99
Electronics (%)                        60.25        61.16
Privatization (%)                       1.54         1.44
N                                       712          623

                                     Hybrid
                                     Auction

                                      Mean
                                    [Median]      t-statistics
                                   (Std. Dev.)   [z-statistics]

Proceeds (millions of NTS)           2853.97       -3.83 ***
                                     [699.38]     [-9.76] ***
                                   (16167.57)
Subscription level (thousand)         593.90       -9.99 ***
                                     [403.05]     [-7.62] ***
                                     (610.77)
Underpricing (%)                       16.31        1 99
                                      [10.80]      [1.05]
                                      (32.04)
Normalized price difference (%)        91.03       -0.74
                                      [71.21]     [-5.67] ***
                                      (86.25)
Normalized underpricing (%)            29.24        0.94
                                      [16.34]      [0.05]
                                      (61.04)
EPS                                     2.42       -2.38 **
                                       [2.06]     [-2.27] **
                                       (1.73)
Book Value                             15.37       -1.94 *
                                      [14.18]     [-1.67] *
                                       (4.98)
Age (year)                             16.04       -0.47
                                      [14.00]     [-1.02
                                       (8.51)
Shares sold (%)                        15.25       -6.99 ***
                                      [14.52]     [-6.11] ***
                                       (7.21)
Adjustment (%)                          3.27        1.81 *
                                      [-0.02]      [1.10]
                                      (23.88)
MKM30 (%)                              -0.96        1.86 *
                                       [0.23]      [1.50]
                                      (10.61)
MKV30 (%)                               1.52        1.87 *
                                       [1.52]      [1.77]
                                       (0.51)
Underwriting fee (%), 1996 to           1.45       -2.67 ***
  1999                                 [1.20]     [-3.25] ***
                                       (0.73)
Lottery processing revenue              2.18        1.46
  (%), 1996 to 1999                    [0.84]      [0.08]
                                       (3.30)
Lottery processing revenue              1.01        1.05
  (%), 2000 to 2003                    [0.85]      [0.09]
                                       (0.90)
TWSE listing (%)                       55.06       -7.57 ***
Electronics (%)                        53.93        1.79 *
Privatization (%)                       2.25       -0.57
N                                       89

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table II. Regressions of Normalized Price Difference and Normalized
Underpricing

This table presents the results obtained from cross-sectional
regressions of the normalized price difference and the normalized
underpricing for IPOs from 1996 to 2003 on the firm characteristics
and market condition. The variable definitions are provided in
Appendix B. The inverse Mills ratios are calculated according to Lee
(1978). Year dummies are included as a control for the time trend.
The t-statistics are reported in parentheses.

                         Normalized Price Difference

Variable              Fixed price offer   Hybrid auction

Intercept                 -0.31             -3.62
                         (-0.28)           (-1.13)
Ln Proceeds                0.05              0.15
                          (0.85)            (1.29)
ELE                        0.21 **           0.51 **
                          (2.25)            (2.45)
TWSE                      -0.24 **          -0.2
                         (-1.99)           (-0.93)
EPS                        0.03              0.06
                          (0.83)            (0.95)
MKM30 (%)                  0.03 ***          0.01
                          (7.04)            (1.56)
MKV30 (%)                 -0.26 **          -0.57 ****
                         (-2.57)           (-2.92)
Adjustment (%) x lO        0.13 ***          0.11 ***
                         (44.89)            (2.69)
Shares sold (%)            2.34 ***          0.37
                          (2.62)            (0.25)
Inverse Mills Ratio       -1.05 *           -1.42
                         (-1.86)           (-1.35)
Year Dummies              Included           Included
N                            623                89
Adjusted [R.sup.2]          0.81               0.36

                           Normalized Underpricing

Variable              Fixed price offer   Hybrid auction

Intercept                 -0.31              -3.23
                         (-0.28)            (-1.32)
Ln Proceeds                0.05               0.13
                          (0.85)             (1.41)
ELE                        0.21 **            0.37 **
                          (2.25)             (2.29)
TWSE                      -0.24 **           -0.08
                         (-1.99)            (-0.46)
EPS                        0.03               0.05
                          (0.83)             (1.14)
MKM30 (%)                  0.03 ***           0.01
                          (7.04)             (1.63)
MKV30 (%)                 -0.26 **           -0.52 ***
                         (-2.57)            (-3.48)
Adjustment (%) x 10        0.03 ***          -0.02
                         (10.10)            (-0.57)
Shares sold (%)            2.34 ***           0.55
                          (2.62)             (0.48)
Inverse Mills Ratio       -1.05 *            -1.24
                         (-1.86)            (-1.54)
Year Dummies              Included           Included
N                            623                89
Adjusted [R.sup.2]          0.27               0.24

** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table III. Comparison of Net Wealth Gain of Pre-IPO Shareholders
if the Alternative Issue Method is Selected

The difference in normalized expected net wealth gain for pre-IPO
shareholders is defined as NWGD = ([PD.sub.f]--PDa)--ss([UP.sub.f]--
[UP.sub.a]), where PDf and PDa are the normalized price differences
in Table II, ss is the ratio of the number of secondary shares sold
in the IPO to the number of all of the shares originally held by
pre-IPO shareholders, and [UP.sub.f] and [UP.sub.a] are the
normalized underpricing in Table II. The projected normalized price
difference (underpricing) of using the alternative issue method is
calculated according to the regression models reported in Table II.
The median value of each variable is reported in parentheses. The
number of sample firms is reported in brackets.

                                                Fixed price    Hybrid
                                                   offer      auction

Panel A: Normalized Price Difference and Underpricing

Normalized price difference (%)                    73.27       91.03
                                                  (21.12)     (71.21)
                                                   [623]        [89]
Normalized underpricing (%)                        40.95       29.24
                                                  (11.04)     (16.34)
                                                   [623]        [89]
Projected normalized price difference when         -9.06       138.88
  using the alternative issue method (%)         (-41.25)     (136.89)
                                                   [623]        [89]
Projected normalized underpricing when            -88.05       135.60
  using the alternative issue method (%)         (-91.18)     (129.90)
                                                   [623]        [89]

Panel B: Expected Net Wealth Gain

Realized net wealth gain (%)                       68.61       86.69
                                                  (19.72)     (69.29)
                                                   [623]        [89]
Projected net wealth gain when using the           0.10        117.12
  alternative issue method (%)                   (-31.60)     (113.21)
                                                   [623]        [89]
Proportion of firms with positive expected         83.31       76.40
  net wealth gain difference when using the        [519]        [68]
  fixed price offer (%)

Table IV. Regressions of Subscription Level

This table presents the results obtained from cross-sectional
regressions of the subscription level for the IPOs from 1996 to 2003
on the firm characteristics and market condition. The variable
definitions are provided in Appendix B. The inverse Mills ratios are
calculated according to Lee (1978). The t-statistics are reported in
parentheses.

Variable              Fixed Price Offer   Hybrid Auction

Intercept                  -4.03 *            4.78
                          (-1.94)            (1.46)
LnProceeds                  0.69 ***          0.54 ***
                           (5.49)            (3.29)
ELE                         1.04 ***          0.40
                           (5.97)            (1.44)
TWSE                       -0.57 **           0.17
                          (-2.39)            (0.58)
EPS                         0.03              0.07
                           (0.39)            (0.75)
Book Value                  0.01             -0.09 ***
                           (0.25)           (-2.78)
MKM20 (%)                   0.10 ***          0.06 ***
                          (10.47)            (3.32)
MKV20 (%)                  -0.49 ***         -0.46 *
                          (-3.00)           (-1.79)
REG                        -1.21 ***         -0.90 ***
                          (-4.27)           (-2.81)
Adjustment (%) x 10         0.02 ***         -0.08
                           (2.71)           (-1.49)
Inverse Mills Ratio         3.12 ***          2.01 **
                           (3.08)            (2.18)
N                            623                89
Adjusted [R.sup.2]          0.31               0.58

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table V. Comparison of Subscription Level and Underwriters' Revenues

In Panel A, the projected subscription level of an IPO using the
alternative issue method is calculated according to the regression
models reported in Table IV. In Panel B, underwriters' revenues are
computed as the total subscription level of a fixed price public
offer tranche multiplied by the lottery processing fee per
application. We normalize underwriters' revenues by the amount of IPO
proceeds. The median value of each variable is reported in
parentheses. The number of IPO firms is reported in brackets

                                               Fixed price    Hybrid
                                                  offer      auction

Panel A: Subscription Level

Realized subscription level (thousand)           197.99       593.90
                                                (113.96)     (403.05)
                                                  [623]        [89]
Projected subscription level when using the      435.69       20.78
  alternative issue method (thousand)           (270.99)      (7.38)
                                                  [623]        [89]
Proportion of firms with an increase in           77.85         0
  subscription level if the alternative           [485]        [0]
  issue method was used (%)

Panel B: Underwriters' Revenues

Realized underwriters' revenues (%)               2.06         1.94
                                                 (0.90)       (0.84)
                                                  [623]        [89]
Projected underwriters' revenues when using       4.84         0.04
  the alternative issue method (%)               (2.78)       (0.02)
                                                  [623]        [89]
Proportion of firms with an increase in the       77.85         0
  underwriters' revenues if the alternative       [485]        [0]
  issue method was used (%)

Table VI. The Determinants of Issue Method

The dependent variable is a binary variable that takes a value of one
when a hybrid auction is selected, and zero when a fixed price public
offer is selected. The variable definitions are provided in Appendix
B. PNWGD is the projected net wealth gain difference between the
fixed price public offer and the hybrid auction estimated in Table
III. PURD is the projected underwriters' revenue difference in the
lottery processing fee between the fixed price public offer and the
hybrid auction estimated in Table V. The TWSE sample only includes
IPO firms listed on the Taiwan Stock Exchange. Year dummies are
included as control for the time trend.

                       (1) Whole Sample           (2) Whole Sample

Variable          Coefficient  z-statistics  Coefficient  z-statistics

Intercept           -12.25      -5.12 ***      -10.66      -3.89 ***
LnProceeds            0.59       5.76 ***        0.52       4.53 ***
EPS                  -0.10      -1.66 *         -0.11      -1.76 *
Privatization        -1.98      -3.75 ***       -1.99      -3.81 ***
PNWGD                -0.92      -2.15 **        -1.04      -2.33 **
PURD                                             0.03       1.13

Year dummies       Included                   Included
N                     712                        712
Pseudo [R.sup.2]     0.42                       0.42

                       (3) TWSE Sample

Variable          Coefficient  z-statistics

Intercept            -8.79       -1.86 *
LnProceeds            0.52        2.68 ***
EPS                  -0.20       -2.15 **
Privatization        -2.06       -3.32 ***
PNWGD                -2.06       -2.30 **
PURD                  0.05        0.97

Year dummies       Included
N                     186
Pseudo [R.sup.2]     0.44

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table VII. Comparison of Underwriters' Retention Risk
for Fixed Price Public Offer and Hybrid Auction

Underwriters in Taiwan have been required to retain 10% to 25% of the
IPO shares for their own purchases and to retain 50% for the
offerings of special or risky high-technology firms since December
1999. The sample in Panel A consists of IPO firms without the
underwriter retention rate, while the sample in Panel B is composed
of IPO firms with the underwriter retention rate from 2000 to 2003.
The variable definitions are provided in Appendix B. A is the number
of sample firms. The median value of each variable is reported in
brackets. The standard deviation of each variable is reported in
parentheses.

                                   Fixed
                                   price          Hybrid
                             N     offer     N    auction      Test

Panel A: IPO Sample Without the Underwriter Retention Rate

Adjustment (%)              71       1.14    46     7.08     -0.96
                                   [-2.01]         [3.15]   [-2.15] **
                                   (38.21)        (21.24)
Offer price adjustment      71       1.14    46    55.11     -7.6 ***
  (%)                              [-2.01]        [52.36]    -7.20 ***
                                   (38.21)        (35.74)
The proportion of firms     32      45.07    42    91.30     -5.07 ***
  with positive offer
  price adjustment (%)
Traditional initial         71      16.29    46    25.72     -1.61
  return (%)                       [10.00]        [17.91]    -2.46 **
                                   (32.05)        (29.06)
EPS                         71       1.89    46     2.61     -1.99 **
                                    [1.56]         [2.08]    -2.90 ***
                                    (1.85)         (2.00)

Panel B: IPO Sample with the Underwriter Retention Rate, 2000 to 2003

Adjustment (%)              391     32.70    13   -12.26      0.98
                                    [3.31]        [-7.06]    [2.00] **
                                  (165.52)        (16.20)
Offer price adjustment      391     32.70    13    14.00      0.41
  (%)                               [3.31]        [16.31]   [-0.98]
                                  (165.52)        (24.81)
The proportion of firms     209     53.45    9     69.23     -1.12
  with positive offer
  price adjustment (%)
Underwriter retention       391     20.35    13    22.23     -1.08
  rate (%)                         [25.00]        [25.00]    -0.96
                                    (6.16)         (5.60)
Traditional initial         391     28.87    13    68.76     -2.50 **
  return (%)                       [11.11]        [43.55]    -1.68 *
                                   (54.98)        (94.52)
EPS                         391      2.10    13     2.93     -2.03 **
                                    [1.86]         [2.98]    -1.86 *
                                    (1.45)         (1.70)

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.
COPYRIGHT 2015 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:initial public offering
Author:Chen, Hsuan-Chi; Wu, Sheng-Ching
Publication:Financial Management
Article Type:Report
Geographic Code:9TAIW
Date:Dec 22, 2015
Words:16835
Previous Article:Extrapolation errors in IPOs.
Next Article:Liquidity benefits from IPO underpricing: ownership dispersion or information effect.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters