SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

We examine the efficiency of initial public offering (IPO) pricing using a sample of over 300 equity carve-outs from 1985 to 2009. The partial adjustment theory posits that the initial return of IPOs is predictable based on private information, but public information is fully incorporated. Prospect theory is consistent with both private and public information not being fully incorporated in the offer price. Our analysis confirms that both price update and initial return of carve-out IPOs can be predicted based on the parent firm's returns during the prepricing and preissuing periods. Further, postissue ownership of the parent firm is associated with significantly higher price update and initial return, while IPOs where the majority of the proceeds are paid out register lower initial return. The size of the subsidiary and relative size of the offering are also significantly related to price update and initial return. These findings are consistent with prospect theory.

Despite extensive research, the underpricing of initial public offerings (IPOs), where the IPO average initial returns tend to be positive and economically significant, remains an interesting area of research. Studies regarding IPO pricing in different countries under various regulatory environments document that issuing firms systematically underprice IPOs relative to the price the market is willing to pay (Ritter, 2003). In a sample of 6,391 IPOs from 1980 to 2003, Loughran and Ritter (2004) report an average initial return, the percentage change between the offer price and the first closing price, of 6.3% with wide variation over time. The smallest initial return (1.9%) is from 1980 to 1989. The largest initial return (32.3%) is from 1999 to 2000.

Our objective is to explore the IPO price-setting process, specifically, how underwriters treat public information throughout the pricing period, with a comprehensive data set of equity carve-outs (ECOs) to gain new insights on the efficiency of IPO pricing. An ECO is the IPO of a portion of a wholly owned subsidiary's common stock. The unique feature of ECOs is that the parent firm continues to hold a controlling interest in the subsidiary after the IPO. The extant literature has examined the impact of private and public information during the registration period on IPO prices. Benveniste and Spindt (1989) posit that since the number of shares allocated to investors in an IPO is apportioned, underwriters reward investors who truthfully reveal their private information during the book-building period by partially incorporating this information into the offer price.1 The more favorable the private information, the higher the price update (the percentage difference between the final offer price and the midpoint of the initial price range) and, proportionately, the greater the initial returns are. Thus, partial adjustment theory predicts that the initial return is positively related to the price update. Hanley (1993), Bradley and Jordan (2002), and Thompson (2010) report significant supportive evidence for this hypothesis.

Underwriters need not compensate investors for providing public information since this information is freely available to all market participants including underwriters and issuers. As such, Benveniste and Spindt's (1989) model predicts that the offer price should fully incorporate all public information available prior to and during the registration period. In essence, the initial return cannot be predicted based on publicly observable information about the IPO's characteristics and market movements. Contrary to this notion, Loughran and Ritter (2002), Bradley and Jordan (2002), and Lowry and Schwert (2004) document a significantly positive relationship between initial return and the return on the market index prior to the offer day, and several firm and issue characteristics including information that is available well before the final offer price is set. This evidence implies that underwriters do not fully incorporate public information into the offer price. That public information during the registration period can be used to predict IPO's initial return implies that the IPO pricing process is not efficient.

Underpricing is costly to both underwriters and issuers. For underwriters, the loss due to underpricing may possibly be offset by the reputational capital of a successful issue and the prospect of future business with satisfied clients. More intriguing is why issuers are not upset about leaving “money on the table” (defined by Loughran and Ritter, 2002, as the product of the number of shares issued in the IPO times the change in the share price of the issuing firm over the first trading day). Loughran and Ritter (2002) invoke prospect theory to develop a behavioral explanation for this phenomenon. In an IPO, while issuers sustain significant losses due to underpricing of the shares sold, they enjoy large gains from the price appreciation of the shares retained. Prospect theory postulates that if issuers integrate the loss due to underpricing of shares sold with the perceived gain from the higher postissue value of the shares retained, they would not be upset about leaving money on the table. Under this scenario, neither private nor public information is fully incorporated in the offer price. Support for the model is derived from Loughran and Ritter's (2002) finding that IPOs that leave a significant amount of money on the table (i.e., more underpriced) are those where both the offer price and the market price are higher than had been originally anticipated.

In Benveniste and Spindt (1989), the high initial return is the equilibrium compensation to investors for revealing private information during the filing period. Loughran and Ritter's (2002) behavioral model does not distinguish between private or public information. For both models, rationing of IPO shares is a critical element. In an environment where IPO shares are not rationed, informed investors have less incentive to reveal private information during the book-building period. Indeed, according to Lowry and Schwert (2004), the significant correlation between initial return and public information is not conclusive evidence against the efficiency of IPO pricing because in a regular IPO, investors cannot trade at the initial or offer prices. In essence, while the initial file range may not represent the true value of the IPO, investors cannot profit from the mispricing.

In an ECO, since the parent firm continues to hold a controlling stake in the carved-out subsidiary after the IPO, the value of the subsidiary is reflected in the price of the parent firm's publicly traded, unrationed shares (Prezas, Tarimcilar, and Vasudevan, 2000; Thompson, 2010). Therefore, despite the rationing of IPO shares, investors can capture the benefits of ownership of subsidiary shares by trading the parent firm's shares in the open market during the book-building period. In this environment, for efficient pricing of subsidiary IPOs, the information contained in the return of the parent firm prior to the initial pricing period should be incorporated in the initial price range, and information contained in the return of the parent firm prior to and during the book-building period should be incorporated in the final offer price. Thus, price update should not be related to the parent firm's returns prior to initial pricing, and the initial return will not be related to the parent firm's returns prior to initial pricing and during the book-building period. Evidence presented contrary to these predictions is inconsistent with the efficiency of the IPO pricing process. In essence, the special features of ECOs afford us the opportunity to shed new insight on the efficiency of IPO pricing.

To examine the efficiency of the pricing of carve-out IPOs, we analyze a sample of over 300 ECOs in the United States over a 25-year period from 1985 to 2009. Parent firms frequently announce the decision to carve out subsidiaries ahead of filing the issuing prospectus with the Securities and Exchange Commission (SEC). We analyze the IPO pricing process in two stages: 1) the announcement of a carve-out to setting the initial price range (prepricing period) and 2) setting the initial price range to the offer day (preissuing period). Since the parent firm's stock, which reflects the valuation of the subsidiary, trades throughout the pricing period, pricing efficiency implies that the ECO's price update should not be related to parent firm's returns during the prepricing period, and the initial return should not be related to parent firm's returns during the prepricing and preissuing periods. Contrary to this notion, our analyses reveal that both price update and initial return are significantly related to the returns of the parent firm during the prepricing and preissuing periods. The market index return over the prepricing and preissuing periods is also a significant determinant of price update and initial return, respectively. In addition, similar to Bradley and Jordan (2002), Lowry and Schwert (2004), and Thompson (2010), we find that carve-out IPO offer prices do not fully reflect public information about observable issue characteristics including offer size, initial price range, underwriter reputation, and the proportion of subsidiary shares retained by the parent firm. We interpret this as compelling evidence against efficiency in IPO pricing.

This paper makes important contributions to the literature regarding IPO pricing. Bradley and Jordan (2002) characterize the underadjustment of IPO prices to public information as a puzzle deserving of further enquiry. Thompson's (2010) conclusion that initial return of carve-out IPOs can be predicted based on the public information available before the offer day corroborates the evidence. However, as Lowry and Schwert (2004) note, these findings do not necessarily imply inefficiencies in IPO pricing. Specifically, because investors cannot buy shares at the filing or offer prices, the predictability of price update or initial return does not represent a profit opportunity. In an ECO, since the parent firm continues to hold a controlling stake in the subsidiary after the IPO, investors have the opportunity to participate in the offering during the pricing period by trading parent firm shares. Our analyses reveal that public information is not fully incorporated in the initial price range or the offer price of carve-out IPOs. As such, the efficiency of IPO pricing continues to be an unresolved issue.

Further, as noted by Lowry and Schwert (2004), while there is no dearth of research on IPOs, most of the analyses focus on the determinants of initial returns. To our knowledge, studies that have examined price updates (Hanley, 1993; Lowry and Schwert, 2004; Edelen and Kadlec, 2005) are relatively few. Our analyses corroborate Lowry and Schwert's (2004) conclusion that the implicit assumption in previous studies that the midpoint of the preliminary price range is the best unbiased predictor of the offer price is open to question.

This paper proceeds as follows. We review the previous literature and develop the hypotheses in Section I. In Section II, we discuss the sample and sources of data. Section III provides the results from the univariate analyses, while the results of the multivariate regression models can be found in Section IV. Section V provides our conclusions.

I. Pricing of ECOs—Hypotheses

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

A. Does IPO Pricing Reflect Private and Public Information?

After filing the prospectus, the underwriter launches the road show to promote the issue to potential clients. In the Benveniste and Spindt (1989) model, during the book-building period, underwriters induce informed investors to reveal their private information about the issuing firm through initial orders. To provide an incentive for truthful revelation, underwriters reward informed investors by not adjusting the offer price to the full extent of the privileged information revealed, allowing these investors to earn higher returns on the opening day of the issue.2Benveniste and Spindt (1989) predict that the higher the perceived quality of the issue, the higher the price update and, proportionately, the greater the initial return. Consistent with this notion, Hanley (1993), Cornelli and Goldreich (2001), Ljungqvist and Wilhelm (2002), Bradley and Jordan (2002), and Thompson (2010) provide evidence that initial return is an increasing function of price update.3

In the partial adjustment model, while private information is only partially reflected in the offer price, public information is fully incorporated as public information is freely available to all investors. Contrary to this, Loughran and Ritter (2002) document a significant and positive relationship between initial return and the return on the value-weighted Center for Research in Security Prices (CRSP) index over a period of 15 trading days prior to the offer date. They find that if the market gains (loses) 1% during the preissue period, the offer price is increased by 0.5% (reduced by 1.5%). This finding is also confirmed by Cornelli and Goldreich (2001) and Lowry and Schwert (2004). Bradley and Jordan (2002) and Lowry and Schwert (2004) further demonstrate that IPO offer prices do not fully incorporate information about several publicly observable firm and issue characteristics including the proportion of shares retained, lead underwriter, offer size, and offer price. Thompson (2010) reports similar findings for ECOs.

That IPO offer prices do not fully incorporate public information raises two questions. First, why do issuers leave money on the table?Loughran and Ritter (2002) invoke prospect theory to present a behavioral explanation for this phenomenon. In prospect theory, investors focus on the change in wealth, rather than its level. The issuers sustain a significant loss of wealth due to underpricing of shares sold but enjoy large gains from the price run-up of the shares retained. Prospect theory postulates that issuers integrate the losses sustained due to underpricing with the perceived wealth gains from higher postissue value of the shares retained. If the perceived gain exceeds the actual loss, the issuers would not be upset about leaving money on the table.4 It is noteworthy that the prospect theory explanation does not require either private or public information to be fully incorporated in the offer price. Corroborative evidence that managers behave in accordance with the predictions of prospect theory is provided by Ljungqvist and Wilhelm (2002) who find that institutions that reveal more information during the book-building period are rewarded with higher allocations when such information is favorable. Ljungqvist and Wilhelm (2005) also demonstrate that IPO firms are less likely to switch underwriters in subsequent seasoned equity offerings when behavioral proxies indicate that they are satisfied with the IPO underwriters’ performance.5

Perhaps a more interesting question is whether the finding that offer prices do not fully incorporate public information indicates inefficiency in IPO pricing. According to Lowry and Schwert (2004), a definitive conclusion based on the analysis of regular IPOs is premature since access to the issuing firm's shares is restricted such that investors do not have the opportunity to trade the issuer's stock prior to the IPO to earn abnormal profits.6 The unique environment of ECOs where the value of the parent firm (which holds controlling interest in the subsidiary) reflects the valuation of the subsidiary and investors can trade the parent firm's shares affords an opportunity for new perspective on the issue. In the next section, we discuss the pricing process of carve-outs and develop our hypotheses regarding the relationship between ECO characteristics and parent firm's returns, and price update and initial return.

B. Pricing of ECOs

The pricing process of ECOs during the registration period begins when the parent firm announces the decision to carve-out the subsidiary through a public offering. In our sample, 154 of 303 total cases (51%) were preannounced. Subsequently, the issuing firm files the preliminary prospectus with the SEC. In most cases, the issuing firm reveals the initial price range on the initial filing day. In some cases, the price range is listed in a subsequent filing. We refer to the day when the initial price range is listed as the pricing day. The offer price is usually set after the market closes on the day before the issue day.

There are two distinct periods in the pricing of ECOs during which new information is released to the market: 1) the period between the announcement day and the pricing day (prepricing period) and 2) the period between the pricing day and the issue day (preissuing period). We focus on the extent to which two sets of publicly available information, 1) the unique characteristics of the carve-outs that are known on the announcement or pricing day and 2) the returns on the parent firm's common stock during the prepricing and preissuing periods, are incorporated in the initial file range and the final offer price. After the initial price range is included in the prospectus, the issuer can adjust/amend this price range at any time prior to the issue date by filing an amended S-1 form. Bradley and Jordan (2002) characterize the offer price relative to the midpoint of the initial file range as total price adjustment. For cases where the initial file range is amended, Bradley and Jordan (2002) use the file range amendments to decompose the total price adjustment into two components: 1) the preoffer adjustment defined as the percentage difference between the midpoint of the final file range and the midpoint of the initial file range and 2) the final adjustment representing the percentage difference between the offer price and the midpoint of the final file range. Thompson (2010) extends the analysis to ECOs. Both Bradley and Jordan (2002) and Thompson (2010) report that initial return is an increasing function of price adjustment.

To examine underwriters’ treatment of available information in the price-setting process of carve-outs, we first focus on price update, which represents the total price adjustment defined as the percentage difference between the offer price and the midpoint of the initial filing range.7 Next, we investigate the extent that public information learned during the prepricing and preissuing periods is incorporated into the offer price. Benveniste and Spindt's (1989) model predicts that public information is fully incorporated into the offer price. Loughran and Ritter's (2002) prospect theory explanation predicts that both private and public information are partially incorporated into the offer price. To examine this notion, we focus on the initial return defined as the percentage difference between the offer price and the closing price on the first day of trading.

1. Impact of Parent Firm and Market Return on Pricing of ECOs

The return of the parent firm's stock during the prepricing period incorporates the market reaction to the carve-out announcement. We define Preturn1 and Mreturn1 as the parent firm's excess return over the CRSP value-weighted index and the return on the CRSP value-weighted index, respectively, from the announcement day to one day prior to the pricing day. Under partial adjustment theory, underwriters incorporate all publicly available information, including the parent firm's return, and the return on the market index during the prepricing period to set the initial file range. As such, there should be no correlation between the return of the parent firm (Preturn1) and the market (Mreturn1) during the prepricing period and price update, and, by implication, the initial return. This leads to our first hypothesis:

  • H1: 
    Under the partial adjustment theory, the parent firm's return and the market return during the prepricing period should not be significantly related to price update and initial return.

The return of the parent firm's stock during the preissuing period incorporates the change in the value of the nontraded subsidiary during this period. Preturn2 and Mreturn2 represent the parent firm's excess return over the CRSP value-weighted index and the return on the CRSP value-weighted index, respectively, from the pricing day to one day prior to the offer day. If the underwriter fully incorporates public information in the offer price, then a positive relationship between the return of the parent firm and the market during the preissuing period and price update is predicted. However, no correlation between the return of the parent firm and the market during the preissuing period and initial return is expected.

  • H2: 
    Under the partial adjustment theory, the parent firm's return and the market return during the preissuing period are significantly related to price update. However, the parent firm's return and the market return during the preissuing period should not be related to the initial return.
2. Impact of Firm and Issue Characteristics on Pricing of ECOs

To focus on the impact of the parent firm's returns on pricing, we must control for firm and deal characteristics. The attributes of ECOs that are known on the announcement and pricing days include the initial file range, the offer size, the relative size of the carved-out subsidiary, the use of IPO proceeds, the diversity of the parent firm's core businesses, and post-IPO ownership of the parent firm. According to partial adjustment theory, if underwriters fully incorporate public information about carve-out characteristics in the initial price range and final offer price, then price update and initial return will not be related to these (observable) characteristics. This leads to our third hypothesis:

  • H3: 
    Since information about issue characteristics becomes available at the announcement and initial pricing of the equity carve-out, price update and initial return should not be significantly related to these observable characteristics if underwriters incorporate public information in IPO pricing.

Our hypotheses are stated from the perspective of partial adjustment theory. Prospect theory implies incomplete incorporation of both private and public information, so violation of the hypotheses is not inconsistent with prospect theory. Previous research has reported a significant correlation between firm and issue characteristics, and both price update and initial returns of regular IPOs (Hanley, 1993; Bradley and Jordan, 2002; Lowry and Schwert, 2004) and ECOs (Prezas et al., 2000; Hogan and Olson, 2004; Thompson, 2010). As Lowry and Schwert (2004) observe, some of the significant associations between price update and initial return, and information known by underwriters prior to the IPO are consistent with efficiency. Benveniste and Spindt (1989) argue that revisions in the initial offer range reflect the information collected by the underwriters during the book-building period such that firms with greater uncertainty surrounding the true value of the shares are more likely to have larger revisions in offer prices and initial returns. Nanda's (1991) conclusion that the parent firm's managers choose a carve-out when they consider the subsidiary assets to be overvalued relative to the parent firm's stock implies that IPO characteristics that mitigate information asymmetry will be related negatively to the initial return.

However, the implications of the asymmetric information model may be tempered by the unique features of carve-outs. First, since the subsidiary trades in the market as a part of the parent firm, investors have access to information about the subsidiary through SEC filings (Prezas et al., 2000). Additionally, because the parent firm continues to own a significant part of the subsidiary after the carve-out, the value of the parent firm's stock incorporates information about the value of the subsidiary (Prezas at al., 2000; Thompson, 2010). As such, information asymmetry is less severe for carve-outs. Consistent with this notion, Prezas et al. (2000), Vijh (2002), and Lowry and Schwert (2004) report that ECOs are less underpriced than regular IPOs. Finally, ECOs enhance value by refocusing parent and subsidiary core businesses, improving managerial incentives through stock-based compensation, and enhancing transparency through independent trading (Schipper and Smith, 1986). Vijh's (2002) analysis leads him to reject the information asymmetry hypothesis in favor of the divestiture gains hypothesis. Powers (2003) also rejects the information asymmetry hypothesis, and concludes that financing is the main objective for an ECO. Overall, due to the disparate implications of the various hypotheses, the influence of firm- and issue-specific variables on pricing of ECOs is an empirical issue. To develop the predicted relationship between firm and deal characteristics and ECO pricing, we review the extant literature below.

C. Offer Size

We use the natural logarithm of expected issue size in millions of dollars. Hanley (1993) hypothesizes that offer size is inversely related to the change in offer price. Lowry and Schwert (2004) argue that investment bankers set the initial price range of riskier issues lower to minimize the chance of the issue being worth less than its projected value. In conjunction with the notion that larger IPOs are easier to value (Booth and Chua, 1996; Prezas et al., 2000), Lowry and Schwert's (2004) argument implies a negative correlation between issue size and price update, and initial return. Bradley and Jordan (2002), Lowry and Schwert (2004), and Edelen and Kadlec (2005) present evidence in support of this hypothesis. However, Prezas et al. (2000) find no relationship between initial return and offer size of ECOs.

D. Relative Size

Relative size is the ratio of offer size to the total assets of the parent firm in the quarter preceding the IPO. Allen and McConnell (1998) and Vijh (2002) report that the parent firm's stock return around the announcement of an ECO is a positive function of relative size. As such, relatively larger divestitures induce greater restructuring gains. Powers (2003) reports that the relative size of ECOs is negatively related to percentage of the subsidiary sold, while Thompson (2010) finds that the initial return is an increasing function of the proportion retained. The evidence jointly leads to the prediction that relative size impacts price update and initial return positively.

E. Offer Depth

We define offer depth as the ratio of the initial file range to the midpoint of the file range. Hanley (1993) and Lowry and Schwert (2004) argue that to retain flexibility in setting the offer price, investment bankers set wider and lower initial price ranges for riskier firms. Hanley (1993) reports a significantly positive association between the width of the initial offer range and absolute price update. Jegadeesh, Weinstein, and Welch (1993), Schultz (1993), Booth and Chua (1996), and Beatty and Welch (1996) argue that offer price is a proxy for uncertainty about value. As offer price increases, price update and initial return decrease. Accordingly, we expect offer depth to be positively related to price update and initial return. However, the significantly positive correlation between initial return and offer price in Prezas et al. (2000) is contrary to this notion.

F. Share Overhang

Bradley and Jordan (2002) refer to the unsold shares relative to the number of shares offered in the IPO as share overhang and observe that only the shares that are sold are underpriced, while shares that are retained are valued at market. We define share overhang as the ratio of pre-IPO shares retained by the parent firm to the number of subsidiary shares filed for sale. Ljungqvist and Wilhelm (2003) speculate that insiders with large stakes bargain for more aggressive price revision. Vijh (1999) posits that because of the parent firm's continued ownership of a controlling interest in the subsidiary, its role is similar to that of a venture capitalist (VC). By providing certification and monitoring, a VC reduces information asymmetry, implying a negative effect on price update and initial return (Megginson and Weiss, 1991). More recently, researchers report a nonnegative effect of VC backing on initial return (Lowry, Officer, and Schwert, 2010). Overall, extant evidence does not yield a definitive correlation between overhang and price update. Regarding the impact of overhang on initial return, Bradley and Jordan (2002) argue that since only the shares sold are underpriced, dilution declines as overhang rises. Therefore, greater overhang alleviates the adverse effect of underpricing. Similarly, prospect theory implies that managers retain more shares when they expect the IPO to be highly underpriced. Accordingly, we predict a positive relationship between initial return and share overhang (Bradley and Jordan, 2002; Loughran and Ritter, 2004). The correlation between post-IPO ownership and initial return is expected to be similar. Hogan and Olson (2004) and Thompson (2010) report results consistent with this hypothesis.

G. Underwriter Rank

Carter and Manaster (1990) developed underwriter rankings based on their relative order in tombstone advertisements, which Carter, Dark, and Singh (1998) later updated. We use the Loughran and Ritter (2004) revised rankings ranging from 1.1 to 9.1 with higher numbers representing underwriters of greater reputation. Assuming that smaller and inexperienced underwriters are more likely to misprice an issue, a negative correlation between underwriter rank and price update is predicted (Hanley, 1993). Conversely, larger, more seasoned underwriters sell to more informed investors who demand larger price revisions. Highly ranked underwriters are also likely to set the initial price range low to avoid the possibility of having to adjust the final offer price down (Lowry and Schwert, 2004). Hanley (1993), Lowry and Schwert (2004), Ljungqvist and Wilhelm (2003), and Edelen and Kadleck (2005) report a significantly positive impact of underwriter rank on price update. The evidence regarding the impact of the quality of the lead underwriter and initial return is mixed. The pre-1990 evidence is consistent with the perception that high-ranking underwriters carry certification value, thereby inducing lower initial returns (Beatty and Welch, 1996). During the post-1990 period, Beatty and Welch (1996) report a positive association between underwriter ranking and initial return, as do Bradley and Jordan (2002), Ljungqvist and Wilhelm (2003), Cliff and Denis (2004), Edelen and Kadlec (2005), and Lowry et al. (2010). Conceivably, underwriters accepted riskier issues to take advantage of the hot market.8 For ECOs, Prezas et al. (2000) and Thompson (2010) find a negative effect of underwriters on initial returns from 1986 to 1995 and 1988 to 1998, respectively. Hogan and Olson (2004) report a positive correlation for the period 1990-2000.

H. Use of Proceeds

We include an indicator variable, Dpayout, which takes a value of one if more than 50% of the proceeds are paid out and zero otherwise. Allen and McConnell (1998) find that the parent firm's stock price response to an announcement of carve-outs is significantly greater when the funds are used to redeem debt or pay dividends. The authors attribute this result to a reduction of agency cost of managerial control over discretionary capital. This predicts a positive association between the payout dummy and both price update and initial returns. Contrary to this notion, Thompson (2010) reports a significantly negative effect of the percentage of offer proceeds paid out to security holders on the initial return of carve-outs.

I. Restructuring

We include an indicator variable, Dfocus, which takes a value of one when the parent and the subsidiary have different Standard Industrial Classification (SIC) codes. Schipper and Smith (1986) and Vijh (2002) report that focus-enhancing carve-outs induce greater parent firm returns, reflecting gains from refocusing and the elimination of cross-subsidies between the parent and the subsidiary. Thompson (2010) predicts that the dummy variable indicating cross-industry carve-outs should be positively related to subsidiary returns, but finds no significant results. In contrast, prospect theory implies that since the parent firm tends to retain a smaller fraction of the carved-out subsidiary in refocusing motivated carve-outs, they will be less underpriced.

In addition to firm and issue characteristics, the extant literature consistently documented a significantly positive relationship between price update and initial return (Hanley, 1993; Bradley and Jordan, 2002; Hogan and Olson, 2004; Thompson, 2010). This is consistent with the partial adjustment theory assertion that the final offer price is only partially adjusted to private information received during the book-building period, and with prospect theory that predicts that neither private nor pubic information is fully incorporated in the offer price.

In Table I, we define the firm- and issue-specific variables and their predicted relationship with price update and initial return. The predicted signs are based on theory and, in most cases, the evidence for regular IPOs and the extension thereof to carve-outs. In the table, “none” implies that theory predicts no relationship between the variables, while “unspecified” suggests that the theory makes no clear prediction. Under partial adjustment theory, publicly available information regarding firm and deal characteristics are fully incorporated in the initial price range and offer price, and is predicted to have no effect (“none”) on price update and initial return. To our knowledge, the correlation between price update and firm- and issue-specific variables of carve-outs is unexplored. However, in light of the strong extant evidence of a positive association between price update and initial return, we maintain the same predicted sign for both in cases where there is some ambiguity. We expect the empirical analysis that follows to shed new light on this issue.

Table I. Variable Definitions and Predicted Relationship with Price Update and Initial Return of ECOs This table provides definitions of the variables used in our analysis and their predicted relation to price update and initial return, based on the partial adjustment theory and prospect theory. “None” implies that the theory predicts that there is no relationship between the two variables. “Unspecified” implies that the theory makes no prediction.
VariableDefinitionPredicted Relation with Price UpdatePredicted Relation with Initial Return
Partial adjustment theoryProspect theoryPartial adjustment theoryProspect theory
  1. aRelationship predicted based on the previous literature as discussed in the text that is not inconsistent with prospect theory. As prospect theory does not differentiate between public and private information, it does not explicitly provide prediction on the relationship between these issue characteristics and initial returns.

Price updatePercentage change between the midpoint of the initial file range and the final offer pricePositivePositive
Initial returnPercentage change between the offer price and the initial trading day closing pricePositivePositive
Issue Characteristics (Information Available on the Announcement or Filing/Pricing Day)—Hypothesis 1
Share overhangRatio of subsidiary shares retained by the parent relative to the number of shares filed for saleNonePositiveNonePositive
Post-IPO ownershipPercentage of the subsidiary shares the parent firm owns after the ECONonePositiveNonePositive
Offer sizeThe natural log of ECO issue size in millions of dollarsNoneNegativeaNoneNegativea
Offer depthRatio of the initial file range to the midpoint of the file range (Hanley, 1993)NonePositiveaNonePositivea
Relative sizeRatio of ECO issue size to the total assets of the parent in the quarter preceding the IPONonePositiveaNonePositivea
Underwriter rankUnderwriter's rank ranging from 1.1 to 9.1 with higher numbers representing underwriters with higher reputation (Loughran and Ritter, 2004)NonePositiveaNonePositivea
DfocusIndicator variable equal to one if the subsidiary and the parent firms have different two-digit SIC codes, and zero otherwiseNoneNegativeNoneNegative
DpayoutIndicator variable equal to one if more than 50% of the proceeds from IPO are paid out, and zero otherwiseNonePositiveNonePositive
Information Available Before Pricing Day—Hypothesis 2
Preturn1Return on parent firm over the CRSP value-weighted index from the announcement date of carve-out to one day prior to the pricing dateNoneUnspecifiedNoneUnspecified
Information Available Before Issuing Day—Hypothesis 3
Preturn2Return on parent firm over the CRSP value-weighted index from the pricing day to one day prior to the issue dayPositivePositiveNonePositive

II. Sample and Data

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

We obtain an initial sample of carve-outs from the Securities Data Corporation's (SDC) database by identifying all publicly traded firms in the United States that sold part of their subsidiary to raise capital from 1985 to 2009. Unit offerings, rights issues, and IPOs that cannot be verified from other sources (i.e., Wall Street Journal, Investors Daily, etc.) are dropped. The initial data contain 452 carve-outs. We require that the parent firm retains at least 50% ownership in the subsidiary after its IPO. Since our objective is to examine if (publicly observable) performance of the parent stock is incorporated in the pricing of the carve-out, our research design requires that the parent firm has traded stock throughout the IPO process starting from the initial announcement. We also require that return data are available in CRSP. This reduces our sample to 267 observations. We eliminate multiple filings thereby reducing the sample by 22 observations. Multiple filings appear in the SDC database when the issuer sells the issue in the United States and a foreign market. The issue date is the same, but the filing dates differ. We only keep the filings in the American market, as we do not have price data available for the foreign markets. In addition, we conduct an independent search in Factiva for ECOs and find 58 announcements that are not contained in the SDC Platinum database for which we are able to track a registration statement and confirm that the parent retained at least 50% ownership in the subsidiary carved out. This brings our sample to 303 observations. Our sample size compares favorably with the sample size of 140 carve-outs from 1991 to 1997 in Frank and Harden (2001), 251 carve-outs from 1986 to 1995 in Prezas et al. (2000), and 271 cases from 1988 to 2006 in Thompson (2010). Our sample is smaller than that of the 458 observations from 1990 to 2000 in Hogan and Olson (2004). However, their research design does not require the parent firm to have traded stock.

For each carve-out, the name of the lead underwriter and the initial file range are collected from the issuing prospectus. For the vast majority (248) of cases, the initial price range was reported in the original SEC filing. For 55 cases, the initial price range was reported in a subsequent S-1 filing. In the rest of the paper, we refer to the day when the initial file range is reported as the pricing day, regardless as to whether it is reported in the original or a subsequent filing. The final offer price and number of shares offered are obtained from the SDC database.

We search the Dow Jones News Retrieval and LexisNexis to identify the day when the ECO is first announced in the news media. Of the 303 observations in our sample, 154 (51%) were announced before the pricing day. Of the 149 that were not preannounced, the announcement day coincided with the pricing day in 87 cases, 59 carve-outs were announced after the pricing day, and we failed to find the announcement for the three remaining cases (two observations in 1985 and one in 1987). The average interval between the announcement day and the pricing day for the cases where we could find announcement dates is 74 days. Finally, the closing price of each subsidiary IPO on the first trading day and trading data for the parent firm during the IPO process are obtained from the CRSP database.

III. Empirical Analysis: Univariate Results

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

The distribution of ECOs over the study period, their characteristics, and the returns on the market index and the parent firm over the prepricing and preissuing periods are presented in Table II, Panels A and B. In Panel A, we find that the market for ECOs was most active over the 10-year period 1991-2001 accounting for 214 (nearly 70%) observations. Over the same period, Thompson (2010) finds 217 observations. The number of offerings peaked around 1999 with 29 observations. Our sample includes only 18 carve-outs post-2001. The slow activity during this period is attributable to the market crash. The average issue size over the study period is $315 million with a discernible increase in issue size from the late-1990s. The average proportion of the subsidiary retained by the parent firm after the IPO is 74%. These numbers compare well with average issue size of $336 million and 73% retention by the parent firm in Thompson (2010).

Table II.  Summary of Characteristics for 303 Carve-Out IPOs from 1985 to 2009 Panel A presents the distribution of our sample by year. Offer size is the ECO issue size in millions of dollars. Post-IPO ownership is the percentage of the subsidiary the parent firm retains after the ECO. Panel B presents summary statistics for our sample for four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). Price update is the percentage change between the offer price and the midpoint of the preliminary price range. Initial return is the percentage change between the first-day closing price and the offer price. Issue size is the ECO issue size in millions of dollars. Relative size is the ratio of ECO issue size to the total assets of the parent in the quarter preceding the IPO. Preturn1 and Mreturn1 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the announcement date to one day prior to the pricing date. Preturn2 and Mreturn2 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the pricing date to one day prior to the issue date. Share overhang is defined as the ratio of the number of shares retained to the number of shares offered in the IPO. Offer depth is calculated as the initial offer range to the midpoint of the initial offer price. Underwriter rank measures the reputation of the underwriter and ranges from 1.1 to 9.1. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Dfocus is an indicator variable equal to one if the IPO and the parent firms have different two-digit SIC codes, and zero otherwise. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO.
Panel A. Distribution of 303 Carve-Outs by Year
YearNo. of ObservationsOffer Size ($M)Post-IPO Ownership (%)
19851249.4668.62
19862037.5475.63
19872077.5874.76
1988740.3473.39
1989942.7867.84
1990468.6877.65
19911966.7566.76
19921358.7165.40
199324198.0468.14
19942358.5869.09
199516214.8975.25
199628292.0278.00
19971499.0476.34
199816835.1474.41
199929372.7878.85
200021818.9379.96
2001101,499.5775.82
200231,359.1074.57
20042887.3069.60
20054417.2866.55
20066336.6788.00
200721,939.2096.50
20091720.0096.50
All303315.1074.02
Panel B. Summary Statistics by Periods
 Prebubble PeriodBubble PeriodPostbubble PeriodTotal PeriodDifferences in Means
 (1)(2)(3)(4)between the Time Periods
 No. of Observations = 225No. of Observations = 50No. of Observations = 28No. of Observations = 303
 MeanMedianMeanMedianMeanMedianMeanMedian(2)-(1)(3)-(1)(3)-(2)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Days between filing and issue dates66.1757.0088.5880.50106.57105.0073.6063.0022.41***40.40***17.99
Price update−0.010.000.180.06−0.010.000.020.000.19***0.00−0.19**
Initial return0.090.030.630.260.120.080.180.060.54***0.03−0.51***
Issue size ($M)170.3741.90560.16123.141,040.54333.60315.1065.60389.79***870.17***480.80
Relative size0.140.060.190.060.200.070.150.060.040.060.02
Preturn10.060.040.250.050.120.060.090.050.19**0.06−0.13
Preturn20.090.040.150.110.050.030.100.050.06−0.04−0.10
Mreturn10.050.040.060.07−0.010.010.040.040.01−0.06**−0.07**
Mreturn20.030.020.050.04−0.010.020.030.020.02−0.04*−0.06**
Share overhang4.233.179.545.298.603.885.523.705.31***4.37***−0.94
Offer depth0.120.130.090.120.120.120.120.13−0.03**0.000.03
Underwriter rank7.488.108.419.108.819.107.768.750.931.33***0.40
Dpayout0.490.000.430.000.541.000.480.00−0.060.050.11
Dfocus0.641.000.521.000.541.000.611.00−0.12−0.110.02
Post-IPO ownership0.720.730.790.830.790.800.740.750.07***0.06***−0.01

In Panel B, we present the univariate statistics over four periods: 1) the prebubble period (1985-1998) with 225 observations, 2) the bubble period (1999-2000) with 50 observations, 3) the postbubble period (2001-2009) with 28 observations, and 4) the total period (1985-2009). These periods compare with Hogan and Olson (2004) who analyzed the pricing of ECOs over two periods, 1990-1998 and 1999-2000, the latter being commonly referred to as a “hot period.” These periods also correspond with the two periods in Thompson (2010), the nonbubble period (1988-1998 and 2001-2006) and the bubble period (1999-2000). The average initial return over our study period is slightly over 18% compared to 19% in Thompson (2010). The average relative size of the carve-out firm is 15% of the value of total assets of the parent. The issue size of nearly 30% of the carved-out subsidiaries exceeds 10% of the value of the parent. This indicates that the subsidiary is typically a significant part of the parent firm. The motivation for over 60% of the carve-outs is to refocus the business of the parent firm. Over the entire period, nearly 50% of the carve-outs paid out more than 50% of the IPO proceeds.

The difference in means analyses reveals that the values of several variables are significantly different during the bubble period. The average issue size (expected proceeds) during the bubble period is significantly larger than that of the prebubble IPOs. The difference between the issue size during the bubble and prebubble periods is $380 million and is significant at the 1% level. Thompson's (2010) sample confirms a similar pattern. The average registration period is more than 20 days longer during the bubble period, an indication of the greater uncertainty of these issues. Carve-out issues are considerably more underpriced in the bubble period with a difference in the initial returns during the two periods of 54%, which is significant at the 1% level and consistent with the other trends of share overhang and price update being significantly higher in this period. Indeed, the correlation between price update and initial return in our sample is 0.51, which is significant at the 5% level. This is consistent with the partial adjustment model. The high initial return during the bubble period has been documented by Loughran and Ritter (2004) for regular IPOs and Hogan and Olson (2004) for ECOs. Loughran and Ritter (2004) contribute the increased initial return to what they call a “changing issuer objective function” that refers to the increasing importance over time, and especially during the bubble period, of analyst coverage and issuers’ willingness to accept a lower price in exchange for research coverage and analyst support. Hogan and Olson (2004) report evidence that supports the changing issuer objective function for ECOs. Consistent with this notion, in our sample, IPOs are associated with significantly higher rated underwriters during the bubble period.

Our analyses also confirm that the high initial return during the bubble period is accompanied by high post-IPO ownership or retention of a larger proportion of subsidiary shares by the parent firm. For example, retention after the carve-out and initial return are at nearly their highest levels in 1999 and 2000.9 The difference between postownership in the prebubble period and the postbubble period is 7% and is significant at the 1% level. We observe a similar trend for the difference in the percentage of shares retained between the bubble and the postbubble period. Bradley and Jordan (2002) and Thompson (2010) report similar findings with share overhang.

A. Strategic Attributes of Carve-Outs

Extant evidence reveals that in sharp contrast to the valuation effect of capital-raising activities, announcements of ECOs induce significantly positive abnormal returns for parent firms (Schipper and Smith, 1986; Allen and McConnell, 1998; Hulburt, Miles, and Woolridge, 2002; Vijh, 2002). Three hypotheses have been advanced to explain this result. The divestiture hypothesis (Schipper and Smith, 1986) asserts that an ECO is a strategy to create a pure play stock for the subsidiary. The separation from the parent facilitates refocusing of the parent and subsidiary operations, financing of the subsidiary's projects, retirement of the parent firm's debt, and a more efficient contract for subsidiary's managers. Vijh (2002), Hulburt et al. (2002), and Boone, Haushalter, and Mikkelson (2003) present evidence supporting this hypothesis. The asymmetric information hypothesis states that firms carve out subsidiaries that are overvalued by the market (Nanda, 1991). Slovin, Sushka, and Ferraro (1995) present early evidence for this hypothesis. Schill and Zhou (2001) analyze a small sample of 12 Internet subsidiary carve-outs and interpret the phenomenon of directly owned shares trading at a large premium to indirect shares held by the parent to be consistent with the notion of parent companies opportunistically choosing carve-out transactions when the subsidiary is valued richly relative to the parent firm. Michaely and Shaw (1995), however, conclude that asymmetric information between management and outsiders concerning the value of the divested assets has no discernible influence on the carve-out decision. Vijh (1999) also rejects the information asymmetry hypothesis in favor of the divestiture gain hypothesis based on the evidence that the parent firm's gain at the announcement of the carve-out is an increasing function of the relative issue size. Powers (2003) attributes a carve-out decision to the financing motive.

Proposed by Allen and McConnell (1998), the managerial discretion hypothesis asserts that an ECO is a way to raise funds by divesting subsidiary assets rather than selling the parent firm's securities. Since carve-outs reduce the asset base under the parent managers’ control, the positive market reaction of the parent firm is attributable to lower agency costs. However, if the funds raised in the IPO are retained in the business rather than used to retire debt, the potential restructuring benefits are largely compromised. Allen and McConnell's (1998) finding that firms that pay out most of the IPO proceeds are associated with greater announcement period abnormal returns when compared to those firms that retain most of the proceeds bears out this hypothesis. Lang, Poulsen, and Stulz (1995) reach similar conclusions.

If the initial return of IPOs reflects the asymmetry of information about the value of the new issue, then it is intuitive that IPO characteristics that impact the value of the parent firm at the announcement of the carve-out would alleviate the uncertainty about the value of the new issue. As noted earlier, separation from a parent in a different industry allows the subsidiary to pursue strategic opportunities and develop a niche market. Prospect theory implies that larger postoffering ownership by the parent firm is a potential signal that the issue is likely to be more underpriced. Ownership by the parent also enhances monitoring. To explore whether the characteristics of the carve-outs differ systematically along these dimensions, we separate our sample into the following subgroups: 1) carve-outs with different two-digit SIC codes for parents and subsidiaries versus those with the same two-digit SIC code for parents and subsidiaries, 2) carve-outs where the subsidiary accounts for more than 10% of the book value of the parent firm versus those where the subsidiary represents less than 10% of the book value of the parent, and 3) carve-outs where more than 50% of the proceeds from the IPO are paid out versus cases where less than 50% of the proceeds are paid out.

Summary statistics for the subgroups, identified based on one of the three criteria described above, are reported in Table III. In Panel A, we compare the characteristics of the 186 carve-outs that have different two-digit SIC codes from their parent firms and 117 IPOs that have the same two-digit SIC code as the parent firms. The relative sizes of the two samples suggest that more carve-outs are attributable to the refocus strategy. Prospect theory model implies that since the parent firm tends to retain a smaller fraction of the carved-out subsidiary in refocusing motivated carve-outs, they will be less underpriced. Consistent with this intuition, the average values of all the relevant variables (i.e., initial return, price update, and overhang) are lower, albeit not significantly, when the SIC codes are different than when they are not.

Table III.  Summary Statistics of Selected Attributes of ECOs from 1985 to 2009 in Subsamples Based on Refocusing Carve-Out, Relative Size of the Carve-Out, and the Percentage of IPO Proceeds Paid OutIssue size is the ECO issue size in millions of dollars. Initial return is the percentage change between the first-day closing price and the offer price. Share overhang is defined as the ratio of the number of shares retained to the number of shares offered in the IPO. Price update is the percentage change between the offer price and the midpoint of the preliminary price range. Filing date is the date that the firm files the preliminary prospectus with the SEC. Issue date is the date that the IPO starts trading. Underwriter rank measures the reputation of the underwriter and ranges from 1.1 to 9.1. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Offer depth is calculated as the initial offer range to the midpoint of the initial offer price.
Panel A. Carve-Out Samples Separated by Two-Digit SIC Codes of Subsidiary and Parent
VariablesIssue SizeInitialSharePriceDays between FilingUnderwriterPost-IPOOffer
 ($M)ReturnOverhangUpdateand Issue DatesRankOwnershipDepth
Different two-digit SIC code for parent and subsidiary (n= 186)
Mean301.480.175.030.0271.267.630.740.12
Median66.000.063.350.0061.008.750.740.13
Same two-digit SIC code for parent and subsidiary (n= 117)
Mean336.760.206.300.0377.327.970.750.12
Median61.630.054.190.0067.008.880.780.13
Difference in means−35.29−0.03−1.27−0.01−6.06−0.34−0.010.00
Panel B. Carve-Out Samples Separated by Relative Sizes of Carve-Out and Parent
Size of IPO more than 10% of the book value of parent (n= 88)
Mean308.230.215.960.0572.187.360.720.12
Median63.810.042.960.0058.508.560.700.13
Size of IPO less than 10% of the book value of parent (n= 215)
Mean317.920.185.330.0174.187.920.750.11
Median65.600.064.140.0063.008.830.770.13
Difference in means−9.690.030.630.05*−2.00−0.56**−0.020.01
Panel C. Carve-Out Samples Separated by the Percentage of IPO Proceeds Paid Out
VariablesIssue SizeInitialSharePriceDays between FilingUnderwriterPost-IPOOffer
 ($M)ReturnOverhangUpdateand Issue DatesRankOwnershipDepth
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

IPOs that pay out more than 50% of the proceeds (n= 127)
Mean429.000.124.870.0180.288.230.740.11
Median96.100.043.210.0067.009.000.750.13
IPOs that pay out less than 50% of the proceeds (n= 136)
Mean214.270.276.560.0575.547.540.750.12
Median56.400.084.600.0067.008.430.770.13
Difference in means214.73***−0.15**−1.68−0.044.740.69***−0.01−0.01

The results for the samples based on the relative size of the subsidiary are reported in Panel B. We observe that initial return is not significantly different among the samples, although price update is significantly different at the 10% level. Further, IPOs of greater relative size are managed by underwriters that are less prestigious as indicated by the underwriter ranking in Loughran and Ritter (2004). The difference in the underwriter's ranking for the two subsamples is significant at the 5% level. Finally, in Panel C, we compare parents that use more than 50% of the carve-out IPO proceeds to pay down debt with those that retain more than 50% of the proceeds. According to Allen and McConnell (1998), paying out the proceeds to retire debt or repurchase shares alleviates the potential agency costs associated with free cash flow at the managers’ discretion. Panel C reports that the “payout” type of ECOs tend to be significantly larger in size and associated with higher ranked underwriters and lower initial return. These differences are significant at the 5% level or better. This is consistent with Thompson's (2010) findings.

B. Single-Variable Regressions

We develop the initial list of variables to be included in the regression analysis based on the evidence on IPO pricing in the previous literature. First, single-variable regressions are estimated for price update and initial return during the prebubble, bubble, postbubble, and total periods. The results are reported in Panel A (price update) and Panel B (initial return) in Table IV. In addition to the variables that are significant in the single-variable regressions, we also report the variables that are not significant in the single-variable regressions, but are significant in the multivariate regression models.

Table IV.  Single-Variable Regression Analysis Determining Price Update and Initial Return in ECOs from 1985 to 2009 This table presents single-variable regression results by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). The dependent variable in Panel A is Price update, the percentage change between the offer price and the midpoint of the preliminary price range. The dependent variable in Panel B is Initial return, the percentage change between the first-day closing price and the offer price. Offer size is the natural log of the ECO issue size in millions of dollars. Relative size is the ratio of the ECO issue size to the total assets of the parent in the quarter preceding the IPO. Preturn1 and Mreturn1 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index, respectively, from the announcement date to one day prior to the pricing date. Preturn2 and Mreturn2 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the pricing date to one day prior to the issue date. Share overhang is defined as the ratio of the number of shares retained to the number of shares offered in IPO. Offer depth is calculated as the initial offer range to the midpoint of the initial offer price. Underwriter rank measures the reputation of the underwriter, and ranges from 1.1 to 9.1. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Panel A. Single-Variable Regression Analysis Determining Price Update in ECOs
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
N2255028303
Mean−0.010.18−0.010.02
(Median)(0.00)(0.06)(0.00)(0.00)
Offer size0.0220***−0.00250.00790.0245***
 (3.63)(−0.08)(0.18)(2.89)
Relative size0.05840.2510−0.06340.0957*
 (1.36)(1.46)(−0.62)(1.79)
Preturn1−0.05880.1331***−0.05790.0629
 (−0.71)(3.84)(−0.42)(0.95)
Preturn20.3619***0.6022*0.54790.4659***
 (4.16)(1.79)(1.12)(3.05)
Mreturn1−0.0196−0.85320.33970.0554
 (−0.06)(−0.91)(0.69)(0.23)
Mreturn21.3601***2.0445**2.3713*1.5970***
 (2.97)(2.25)(2.07)(4.22)
Share overhang0.00020.0023−0.00160.0027**
 (0.16)(1.24)(−0.45)(2.10)
Offer depth−0.2452*−0.06560.4748−0.2547
 (−1.80)(−0.12)(0.38)(−1.44)
Underwriter rank0.00510.0303**0.05670.0115**
 (1.14)(1.99)(0.92)(2.41)
Dpayout−0.0151−0.1419−0.0008−0.0443
 (−0.60)(−1.55)(−0.01)(−1.59)
Post-IPO ownership−0.01491.0067***0.21950.3042**
 (−0.16)(2.71)(0.64)(2.48)
Panel B. Single-Variable Regression Analysis Determining Initial Return in ECOs
Dependent Variable =PrebubbleBubblePostbubbleTotal
Initial ReturnPeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean0.090.630.120.18
(Median)(0.03)(0.26)(0.08)(0.06)
Offer size−0.0067−0.1724*−0.01160.0079
 (−0.66)(−1.91)(−0.47)(0.69)
Relative size0.13110.2423−0.1718*0.1648
 (1.02)(0.55)(1.70)(1.18)
Preturn10.00730.4168***−0.2075*0.2698
 (0.07)(3.33)(−1.99)(1.64)
Preturn20.4048***0.5875*0.44110.4967***
 (3.24)(1.72)(1.46)(3.25)
Mreturn10.0720−0.59310.89210.2809
 (0.29)(−0.23)(1.66)(0.67)
Mreturn20.9219**3.93331.75901.4654***
 (2.22)(1.45)(1.35)(2.64)
Share overhang0.00370.00020.00160.0062***
 (0.80)(0.06)(0.31)(2.63)
Offer depth−0.3987**0.6880−0.2162−0.4742
 (−1.98)(0.38)(−0.82)(−1.17)
Underwriter rank−0.00990.06920.01600.0105
 (−0.96)(1.23)(0.20)(0.90)
Dpayout−0.0015−0.6309**−0.1088−0.1507**
 (−0.04)(−2.63)(−1.15)(−2.45)
Post-IPO ownership−0.12391.9154***0.15070.6166***
 (−0.73)(3.01)(0.44)(2.81)

In Panel A, we report the results for price update. We find significant correlation between price update and Preturn1 in the bubble period and Preturn2 during the prebubble, bubble, and total periods. In the prebubble period, Preturn2 has a t-statistic of 4.16 and is significant at the 1% level. In the bubble model, Preturn2 has a lower t-statistic of 1.79 and is significant at the 10% level. However, Preturn1 has a t-statistic of 3.86 and is significant at the 1% level. In the total period, Preturn2 remains significant at the 0.01 level. Market return (Mreturn1) has no effect on price update, but Mreturn2 has a significantly positive impact during all periods. During the total and prebubble periods, the coefficient on Mreturn2 is significant at the 1% level, while during the bubble and postbubble periods its significance is lower at the 5% and 10% levels, respectively. This evidence is inconsistent with the partial adjustment theory and efficiency of IPO pricing, but is not inconsistent with prospect theory.

Contrary to our expectations and the evidence in Lowry and Schwert (2004), we find a significantly positive impact of offer size for the prebubble and total periods, and for relative size for the total period. The coefficient of offer size is significant at the 1% level in the prebubble and total periods, while relative size is significant at the 10% level in the total period. We find a weakly negative effect of offer depth during the prebubble period. Consistent with our prediction, share overhang is significant at the 1% level and positive in the total period, and post-IPO ownership is significantly positive at the 1% and 5% levels during the bubble and total periods, respectively. The strong positive effect of underwriter rank, significant at the 5% and 1% levels in the bubble and total periods, respectively, is consistent with Edelen and Kadlec (2005) and Lowry et al. (2010), confirming the conjecture that highly ranked underwriters set the initial price range low to avoid having to adjust the final offer price down. The indicator variable for payout has no effect on price update.

The results for initial return are reported in Panel B. The return of the parent firm during the prepricing period (Preturn1) is positive and significant at the 1% level in the bubble period and negative and significant at the 10% level in the postbubble period. However, the return of the parent firm during the preissuing (Preturn2) period is positive and significant at the 10% level or better in the prebubble, bubble, and total periods. The preissuing period market return (Mreturn2) is positive and significant at the 5% level or better in the prebubble and total periods. This evidence is inconsistent with the partial adjustment theory and violates IPO pricing efficiency.

Consistent with our hypothesis, the significantly negative effect of offer size during the bubble period corroborates the evidence in Edelen and Kadleck (2005). For carve-outs, however, Prezas et al. (2000) report an insignificant effect for offer size. We find a negative, marginally significant relation between relative size and initial return during the postbubble period. Contrary to our prediction, offer depth is negatively related to initial return during the prebubble period. This evidence is consistent with Prezas et al. (2000) who report a significantly positive correlation between offer price and initial return, but contrary to Bradley and Jordan (2002) who find a positive association between initial return and the reciprocal of the original midrange file price. Consistent with our expectations, Hogan and Olson (2004), and Thompson (2010), share overhang is significant at the 1% level during the total period, while post-IPO ownership is significant at the 1% level over the bubble and total periods. This result supports the prospect theory explanation that the parent firm retains more subsidiary shares for issues that are highly underpriced such that the loss on underpricing is offset by the perceived gains on the shares retained. Underwriter rank is not significant in any period. This is in contrast with Prezas et al. (2000) and Thompson (2010) who report a negative impact on underwriter reputation and Hogan and Olson (2004) who report a significantly positive coefficient. The significantly adverse impact of the indicator variable on payout for the bubble and total periods resonates with the similar effects of the percentage of offering proceeds used for debt reduction or to pay dividends in Thompson (2010).

C. Summary—Univariate Analysis

The univariate analysis reveals several patterns that are similar to evidence received on more comprehensive samples of IPOs. In support of the partial adjustment theory, the data reveal positive and significant correlations between price update and initial return. The evidence that in carve-outs that are more underpriced, the issuer retains a larger postoffering stake lends support to the prospect theory postulate that issuers offset the loss in underpricing with the gains on retained shares. The data find that in the majority of cases, focusing parent businesses is the main objective. As anecdotal evidence for the premise that the performance of the parent firm reflects the value of the subsidiary, the data reveal that the size of most subsidiary IPOs exceeds 10% of the value of the parent firm's assets.

The analysis further reveals that price update and initial return are significantly related to the parent firm and market returns during the prepricing and preissuing periods, as well as to several firm and issue characteristics. This evidence indicates the inefficiency in pricing of carve-out IPOs. Further, several relationships are contrary to our predictions and to received theory and evidence on the pricing of regular IPOs. Before any definitive interpretations and conclusions are reached, we review the results of our multivariate analysis.

IV. Multiple Regression Analysis

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

A. Impact of Issue Characteristics on IPO Pricing

Next, we estimate heteroskedasticity-corrected multivariate regression models of price update and initial return as functions of observable IPO characteristics (such as offer size, relative size of the subsidiary, proportion of subsidiary shares retained by the parent firm, etc; please see Table I for a list and definitions of issue characteristics examined).10 The results for price update are provided in Table V. As previously reported, price update is, on average, 2% over the period studied. We estimate four models, one each for the prebubble, bubble, postbubble, and the full period with 303 observations. For each period, we begin with the significant variables from the single-variable analysis. To this group, we add two interaction terms (Focus*Retain and Focus*Retain2) between the focus dummy (Dfocus) and the percentage of the subsidiary retained (Post-IPO ownership).11 It is intuitive that to maximize restructuring benefits, the parent firm will retain fewer subsidiary shares when the parent and the subsidiary have different SIC codes. According to the divestiture hypothesis and prospect theory, carve-outs where issuers retain less are associated with lower price update and initial return. Thus, the interaction variable is predicted to have a negative coefficient.

Table V.  Multiple Regression Analysis Determining Price Update in ECOs during 1985-2009 This table presents ordinary least squares (OLS) regression results for the dependent variable, Price update, by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). Price update is the percentage change between the offer price and the midpoint of the preliminary price range. Offer size is the natural log of the ECO issue size in millions of dollars. Relative size is the ratio of the ECO issue size to the total assets of the parent in the quarter preceding the IPO. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Focus*Retain is an interaction variable between Dfocus, an indicator variable equal to one if the IPO and the parent firms have different two-digit SIC codes and Post-IPO ownership. Focus*Retain2 is an interaction variable between Dfocus and Post-IPO ownership squared. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean−0.010.18−0.010.02
(Median)(0.00)(0.06)(0.00)(0.00)
Intercept−0.1081**0.2653−0.0625−0.0439
 (−2.28)(1.49)(−0.32)(−0.87)
Offer size0.0249***−0.00950.01780.0196*
 (3.02)(−0.29)(0.36)(1.80)
Relative SIZE0.0831*0.0447−0.00190.1141**
 (1.75)(0.26)(−0.02)(2.03)
Focus*Retain−0.0777−1.8109***−0.8674−0.5495**
 (−0.48)(−2.93)(−1.08)(−2.42)
Focus*Retain20.12912.2836***0.88640.7090**
 (0.65)(2.79)(1.00)(2.37)
Dpayout−0.0319−0.12600.0097−0.0513*
 (−1.22)(−1.38)(0.07)(−1.69)
Adj. R2 (%)6.3525.898.269.36
F-value2.29**2.39**0.562.95***

To select the final list of variables to be included in the multiple regression model, we apply the Akaike information criterion (AIC) to the initial set of variables after testing for multicollinearity using the variance inflation factor (VIF) measures.12 Next, we select the model with the lowest AIC for each period. For each model, the AIC constitutes a trade-off between explained variance and the number of independent variables. Finally, we include all selected (significant) variables from each of the four models to contrast the significance of coefficients over the four periods.

The models for the prebubble and the periods reveal that carve-outs of larger subsidiaries, both in terms of absolute and relative issue size, are associated with significantly higher price update. This result is contrary to our prediction and in contrast with the significantly negative effect of the amount of offer proceeds on price update of regular IPOs in Lowry and Schwert (2004). Conceivably, due to potentially greater restructuring benefits, larger divestitures induce greater market valuation. The payout dummy is negative and significant for the whole period. The negative effect of the interaction term (Focus*Retain), which is significant at the 1% level, confirms our intuition that focus-enhancing carve-outs where issuers retain less are less underpriced.13 The significantly positive coefficient of Focus*Retain2 implies that the sensitivity of price update to the percentage of ownership retained becomes higher at higher levels of retained ownership.

In Table VI, the dependent variable, Initial return, is, on average, 18% over the period of our study. As in Table V, we estimate four models, one for each period, following the same significance for the prebubble or postbubble periods. Offer size is negative and significant at the 10% level in the bubble period, confirming the notion that larger firms tend to be more transparent and, as such, marked by lower uncertainty and underpricing. This result is consistent with the findings of Carter et al. (1998), Edelen and Kadlec (2005), and Lowry et al. (2010) for US IPOs, and Ejara and Ghosh (2004) for IPOs of American Depositary Receipts (ADR). For the bubble and total periods, carve-outs with larger post-IPO ownership are significantly more underpriced. This is consistent with our prediction and the prospect theory implication that retaining greater ownership after the IPO allows managers to recoup the loss sustained from the underpricing of shares sold. Dpayout has a negative coefficient, which is significant at the 5% level, during the prebubble and total periods, contrary to the notion that paying out of IPO proceeds alleviates agency costs (Allen and McConnell, 1998). Thompson (2010) reports similar results.

Table VI.  Multiple Regression Analysis Determining Initial Return of ECOs during 1985-2009 This table presents OLS regression results for the dependent variable, Initial return, by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). Initial return is the percentage change between the first-day closing price and the offer price. Offer size is the natural log of the ECO issue size in millions of dollars. Relative size is the ratio of the ECO issue size to the total assets of the parent in the quarter preceding the IPO. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean0.090.630.120.18
(Median)(0.03)(0.26)(0.08)(0.06)
Intercept0.04790.21370.2456−0.3114*
 (0.28)(0.25)(0.72)(−1.77)
Offer size−0.0052−0.1874*−0.0087−0.0076
 (−0.40)(1.88)(−0.32)(−0.52)
Relative size0.1584−0.1027−0.15560.1939
 (1.03)(−0.19)(−1.49)(1.36)
Post-IPO ownership0.05822.0392**0.01500.776***
 (0.27)(2.28)(0.05)(3.17)
Dpayout0.0048−0.5042**−0.0937−0.1308**
 (0.12)(−2.17)(−1.04)(−2.18)
Adj. R2 (%)2.4221.159.295.72
F-value0.413.76***0.603.36***

Overall, since carve-out characteristics are publicly observable prior to or on the pricing day, their predictive power for price update and initial return is inconsistent with our hypothesis and contradicts the partial adjustment theory. These findings are consistent with the prospect theory explanation, however.

B. Impact of Parent Firm's Performance on IPO Pricing

The main premise of this study is that since the parent firm owns a controlling stake in the subsidiary, both before and after the IPO, the performance of the parent firm during the prepricing and preissuing periods will reflect the changing value of the subsidiary stock. Although Prezas et al. (2000) and Thompson (2010) note the link between the subsidiary and parent firm performance and valuation, evidence on this issue is scarce. Significant anecdotal support for this link draws from the pattern in our data that, on average, the expected proceeds from the sale of the subsidiary account for as much as 15% of the total assets of the parent firm (Table II). Indeed, except for 1993 and 1996, the size of the subsidiary relative to the parent exceeds 15%, rising as high as 31% in 1994, one of the busiest years. In addition, for nearly 30% of carve-out IPOs in our sample, issue size exceeds 10% of the book value of the parent. In essence, the size of the subsidiary is sufficiently large for the change in its value to have substantial impact on the value of the parent firm.

To provide more direct evidence on this issue, we examine the performance of the subsidiary and the parent stocks over 150 trading days after the carve-out. The correlation between daily returns of the parent and the subsidiary is highly significant (68%). The correlation between the parent and the subsidiary excess returns over the CRSP equally weighted (value-weighted) index is also highly significant at 63% (66%). Next, we regress the excess returns of the parent and the subsidiary against each other. The coefficient is significant at the 1% level with a t-value of 12.68 and R2 of 0.43. In addition, we find that the correlations are increasing in both the proportion of the subsidiary owned by the parent after the carve-out and the increase in focus of the parent. We interpret this as compelling evidence that the performance of the subsidiary is a significant determinant of the valuation of the parent firm.

1. Determinants of Price Update

Next, we examine if the returns of the parent firm and the market during the prepricing (Preturn1, Mreturn1) and preissuing (Preturn2, Mreturn2) periods are fully incorporated in the initial file range and the offer price. As predicted by the partial adjustment theory and Hypotheses 1 and 2, if the parent and market returns are fully incorporated in pricing, there should be no correlation between price update and these returns during the prepricing period and the initial return should be unrelated to the parent and market returns during the prepricing and preissuing periods.

To examine whether public information affects the initial filing range, we examine the relationship between price update and the return of the parent firm and the market when controlling for other firm- and issue-specific characteristics. The results are reported in Table VII. Four regression models, one for each period, are estimated to examine the predictability of price update. The independent variables include those for which public information is available prior to the pricing day, specifically the return on the parent firm over the CRSP value-weighted index and the return on the market index during the period between the announcement day and one day prior to the initial pricing day. As previously mentioned, 154 carve-outs in the sample were preannounced. For these cases, we use the actual period between the announcement day and the day before the initial pricing day. For the remaining cases, we substitute the average number of days between the announcement day and the initial pricing day for the 154 cases, which, as reported earlier, is 74 days.14

Table VII.  Multiple Regression Analysis Determining Price Update Including Parent's and Market's Returns during the Prepricing and Preissuing Periods in ECOs during 1985-2009 This table presents OLS regression results by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). The dependent variable is Price update or the percentage change between the midpoint of the initial filing range and the final offer price. Preturn1 and Mreturn1 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index, respectively, from the announcement date to one day prior to the pricing date. Preturn2 and Mreturn2 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the pricing date to one day prior to the issue date. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Focus*Retain is an interaction variable between Dfocus, an indicator variable equal to one if the IPO and the parent firms have different two-digit SIC codes and Post-IPO ownership. Focus*Retain2 is an interaction variable between Dfocus and Post-IPO ownership squared. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean−0.010.18−0.010.02
(Median)(0.00)(0.06)(0.00)(0.00)
Intercept−0.0857***0.0039−0.0161−0.0659**
 (−2.93)(0.03)(−0.11)(−2.17)
Preturn1−0.02590.2198***0.09360.1109**
 (−0.44)(4.47)(0.54)(2.02)
Mreturn10.1028−0.05080.74900.2331
 (0.42)(−0.07)(0.96)(1.13)
Preturn20.3048***0.4531*−0.54460.3836**
 (3.06)(1.71)(−1.10)(2.41)
Mreturn21.1520**1.7937**3.2038**1.3050***
 (2.54)(1.96)(2.16)(3.17)
Focus*Retain0.1412−1.2840**−1.3882*−0.2868
 (0.66)(−1.93)(−1.81)(−1.19)
Focus*Retain2−0.10611.8811**1.7098**0.4382
 (−0.41)(2.13)(2.17)(1.39)
Adj. R2 (%)23.7562.8557.1429.60
F-value3.81***4.61***2.07*5.02***

To choose the independent variables, we identify the best-fit model based on the results from the stepwise selection using the AIC. It is worth noting that when using the AIC criterion, none of the issue characteristics (e.g., offer size, offer depth, overhang, post-IPO ownership, relative size, and underwriter rank) appears to be significant. One potential explanation is that the impact of these variables is subsumed in the stock returns of the parent firm. The final model includes Preturn1, Preturn2, Mreturn1, and Mreturn2, and the two interaction terms (Focus*Retain and Focus*Retain2) between the focus dummy (Dfocus) and percentage of subsidiary retained (post-IPO ownership).

The regression models, presented in Table VII, are estimated for each period, prebubble (1985-1998), bubble (1999-2000), and postbubble (2001-2009), separately. We observe that Preturn1 is positively related to price update and significant at the 1% and 5% levels in the bubble and whole periods, respectively. This suggests that price update can be predicted by the stock price movements of the parent firm before setting the initial price range. Ostensibly, the underwriters do not fully incorporate public information in the initial price range. Preturn2 is positively related to price update during the prebubble, bubble, and whole periods and is significant at the 10% level or better. This supports the prediction based on the partial adjustment theory that the parent firm's return during the preissuing period is significantly related to price update. However, that Preturn1 still has predictive power for price update during the bubble and total periods confirms that underwriters do not incorporate the information contained in the parent firms’ return in setting the initial price range. Consistent with Lowry and Schwert (2004), Mreturn2 is significant over all periods at the 5% level or better. Finally, the relationship between the interaction variable (Focus*Retain) is negative and significant at the 10% level or better during the bubble and postbubble periods, indicating the market's disapproval of the high retention of subsidiary stock in carve-outs that are focus motivated. However, the effect is nonmonotonic as revealed by the significantly positive coefficient of Focus*Retain2 in the bubble and postbubble periods. These results contradict Hypotheses 1 and 2 and confirm that underwriters do not fully incorporate public information in setting the initial price range.

2. Determinants of Initial Return

Finally, we examine the relationship between initial return and the information content of the parent firm's returns during the prepricing and preissuing periods. The results of the multiple regression analyses are presented in Table VIII. On the basis of the AIC criterion, the model for each period includes Preturn1, Preturn2, Mreturn1, Mreturn2, relative size, post-IPO ownership, and Dpayout. If underwriters incorporate the parent firm's prepricing and preissue returns in the initial file range, there should be no relationship between initial return and Preturn1 and Preturn2. We observe that Preturn1 is positively related to initial return during the bubble and total periods and significant at the 5% level or better, while Preturn2 is positive and significant at the 10% level or better during the prebubble, bubble, and total periods. For the bubble and total periods, both Preturn1 and Preturn2 are simultaneously significant. Furthermore, Mreturn1 is significant and positive during the postbubble period and Mreturn2 is significant and positive during the bubble and postbubble periods. These results are inconsistent with Hypotheses 1 and 2.

Table VIII.  Multiple Regression Analysis Determining Initial Return Including Parent's and Market's Returns during the Prepricing and Preissuing Periods in ECOs during 1985-2009 This table presents OLS regression results by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). The dependent variable is the initial return of the ECO issue, the percentage change between the first-day closing price and the offer price. Preturn1 and Mreturn1 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the announcement date to one day prior to the pricing date. Preturn2 and Mreturn2 are the parent firm excess return over the CRSP value-weighted index and the CRSP value-weighted index return, respectively, from the pricing date to one day prior to the issue date. Relative size is the ratio of ECO issue size to the total assets of the parent in the quarter preceding the IPO. Post-IPO ownership is the percentage of the subsidiary the parent firm owns after the ECO. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean0.090.630.120.18
(Median)(0.03)(0.26)(0.08)(0.06)
Intercept−0.11122.3608−0.5884−0.1589
 (−0.01)(1.68)(−0.86)(−0.93)
Preturn10.01690.7790***−0.02140.2978**
 (0.18)(4.34)(−0.12)(1.99)
Mreturn10.1772−2.73781.1322*0.3286
 (0.71)(−1.35)(1.89)(0.81)
Preturn20.3767**0.5949*−0.12220.5218***
 (2.41)(1.79)(−0.11)(3.58)
Mreturn20.60247.8568***1.96601.3691**
 (1.64)(3.46)(0.55)(2.33)
Relative size−0.0578−5.4891***−1.1799−0.2713*
 (−0.35)(−3.04)(−0.90)(−1.84)
Post-IPO ownership0.0561−1.21721.2279*0.4538*
 (0.33)(−0.87)(1.74)(1.77)
Dpayout−0.0045−1.0285**−0.1326−0.1522**
 (−0.12)(−2.46)(−0.61)(−2.07)
Adj. R2 (%)16.1860.9849.0819.49
F-value2.01*7.83***1.79*3.74***

Relative size has a negative effect on initial return during the bubble and total periods, which is significant at the 10% level or better. This possibly reflects the potential gains from the refocusing of the business following a large divestiture. As predicted, post-IPO ownership is significantly positive in the postbubble and total periods, corroborating our expectation that IPOs where insiders retain more shares are more underpriced. Similar to Thompson (2010), the indicator variable for payout is negative and significant at the 5% level for the bubble and total periods. The predictability of the initial return based on the observable carve-out characteristics clearly violates Hypothesis 3. Overall, we infer that although the results vary by period, they confirm that IPO pricing is not efficient in that underwriters do not incorporate the information contained in the parent firms’ returns and the market's returns, as well as the observable IPO attributes in setting the initial price range or the final offer price.

C. Robustness Check

The previous literature (Hanley, 1993; Bradley and Jordan, 2002) documents a significant relationship between price update and initial return. However, as Table VII demonstrates, price update is not exogenous. It is related to the parent and market index returns and IPO characteristics. Therefore, in Table IX, we estimate a two-stage least squares model including the predicted price update from the first stage, based on the best-fit model from Table VII, as an independent variable in the model to estimate initial return. As in other regression estimates, the best-fit model is chosen by including all the variables from each period according to the AIC.

Table IX.  Multiple Regression Analysis Determining Initial Return Including Predicted Price Update in ECOs during 1985-2009 This table presents OLS regression results by four periods: 1) prebubble (1985-1998), 2) bubble (1999-2000), 3) postbubble (2001-2009), and 4) total (1985-2009). The dependent variable is the initial return of the ECO issue, the percentage change between the first-day closing price and the offer price. Predicted price update is the predicted price update from the selection model based on the best-fit model for each period reported in Table VII. Offer size is the natural log of the ECO issue size in millions of dollars. Relative size is the ratio of ECO issue size to the total assets of the parent in the quarter preceding the IPO. Underwriter Rank measures the reputation of the underwriter, and ranges from 1.1 to 9.1. Offer depth is calculated as the initial offer range to the midpoint of the initial offer price. Dpayout is an indicator variable equal to one if more than 50% of the proceeds from the IPO are paid out and zero otherwise. Standard errors are adjusted for heteroskedasticity. t-Statistics values are reported in parentheses.
Dependent Variable =PrebubbleBubblePostbubbleTotal
Price UpdatePeriodPeriodPeriodPeriod
 (1)(2)(3)(4)
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

N2255028303
Mean0.090.630.120.18
(Median)(0.03)(0.26)(0.08)(0.06)
Intercept0.1232−2.02042.22170.2438**
 (1.64)(−1.23)(1.16)(2.46)
Predicted price update0.9061***1.1566***1.3319*1.0778***
 (3.21)(3.72)(1.73)(3.43)
Offer size0.0026−0.2839***−0.0152−0.0323
 (0.24)(−2.69)(−0.32)(−1.45)
Relative size−0.0760−2.9696**−0.4321−0.2906
 (−0.59)(−2.64)(−0.36)(−1.31)
Underwriter rank−0.00480.4499*−0.2398−0.0080
 (−0.50)(1.95)(−1.00)(−0.58)
Offer depth−0.05914.6566***2.21051.1613*
 (−0.21)(2.77)(0.94)(1.94)
Dpayout−0.0009−0.7808**−0.1551−0.1083*
 (−0.02)(−2.50)(−0.87)(−1.75)
Adj. R2 (%)13.9172.5349.7728.58
F-value1.95*7.53***0.853.02***

We find that consistent with the extant evidence and the partial adjustment theory, during the prebubble, bubble, and total periods, predicted price update is significantly related to initial return. This evidence corroborates the results in Thompson (2010). The significantly negative coefficient of offer size in the bubble period is in agreement with our hypothesis. As in Table VIII, relative size is significantly negative in the bubble period. Consistent with our prediction, underwriter reputation and offer depth are significantly and positively related to initial return during the bubble period. The significantly positive coefficient on offer depth is consistent with the notion that riskier issues are more underpriced. Bradley and Jordan (2002) report similar results for regular IPOs, but Prezas et al. (2000) report opposite results for carve-outs. The positive relationship between investment bankers’ reputation and initial return during the bubble period is consistent with the changing issuer objective function hypothesis (Loughran and Ritter, 2004; Hogan and Olson, 2004). This finding is in agreement with Lowry et al. (2010) who report that over 1981-2005, the correlation between initial return and underwriter reputation is significantly positive, but it disappears if the bubble period is excluded from the analysis. Ejara and Ghosh (2004) report similar results for US IPOs in their study of ADR IPOs. Thompson (2010) also reports a positive, albeit insignificant, relationship between underwriter rank and initial return during the bubble period. It is worth noting that underwriter rank is not significant in the total period model. Finally, issues where more than 50% of the proceeds are paid out are significantly less underpriced during the bubble and total periods. While inconsistent with Allen and McConnell (1998), this result corroborates similar findings by Thompson (2010).

V. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References

In this paper, we examine the efficiency of the pricing process of ECOs, specifically, how underwriters incorporate information regarding the value of the issuing firm as it becomes publicly available into the initial price range and the final offer price. In carve-out IPOs, parent firms, on average, own 74% of the issuing firm's shares after the offerings, and stocks of the parent firms are not rationed and are continuously traded during the preoffering period. As such, the stock price movement of the parent firm should reflect the market valuation of the issuing subsidiary as the news of the offering becomes available during the preoffering period. Since this information is publicly available, IPO pricing efficiency requires that underwriters incorporate it in setting the initial filing range and the final offer price of the carve-out, such that there should be no association between the parent firm's return during the prepricing period and price update. Similarly, no relationship should exist between the parent firm's returns during the prepricing and preissuing periods and the initial return of the subsidiary IPO.

We find that the return of the parent firm during the prepricing period is positively and significantly related to price update and initial return. The parent firm's return during the preissuing period is positively related to price update and initial return, as well. Price update and initial return are also significantly related to the returns of the market during the preissuing period. To further explore the predictability of public information on the pricing of carve-out IPOs, we identify several publicly observable carve-out characteristics that are significantly associated with the excess returns of the parent firm during carve-out announcements. Our analyses reveal that these variables are significantly related to price update and initial return. We compare the pricing process in samples of carve-outs where the parent firm and the subsidiary have the same two-digit SIC codes with those where the parent firm and the subsidiary have different two-digit SIC codes, carve-outs where the subsidiary represents more than 10% of the parent firm's total assets with those where the subsidiary has less than 10% of the parent firm's total assets, and carve-outs where the parent firm pays out more than 50% of the proceeds with those where less than 50% of the IPO proceeds are paid out. The analyses reveal that offer size, the relative size of the subsidiary, payout pattern, and post-IPO ownership by the parent firm are significantly related to price update and initial return. While the relationship between price update and the parent firm's return during the preissuing period is consistent with the partial adjustment theory, based on the rest of the results, we conclude that underwriters do not fully incorporate information regarding the market valuation of the new issue in the initial price range and the final offer price, and that pricing of carve-out IPOs is not fully efficient.

Benveniste and Spindt's (1989) partial adjustment theory provides an explanation as to why private information is not incorporated in IPO pricing, but it maintains that IPO pricing is efficient with respect to public information. Contrary to this, Loughran and Ritter (2002) and Lowry and Schwert (2004) present evidence that market returns during the registration period are not incorporated in IPO pricing. However, as Lowry and Schwert (2004) note, while the finding that underwriters tend to omit available information is puzzling, this does not constitute conclusive evidence against the efficiency of IPO pricing as regular IPO shares are nontraded and rationed so that investors cannot trade at initial or offer prices. In an ECO, since the parent firm's shares, which reflect the value of the subsidiary carved out, are nonrationed and continuously traded, our analyses provide compelling evidence against IPO pricing efficiency.

We close with two final observations. First, Bradley and Jordan (2002) characterize the under-adjustment of IPO prices to public information as a puzzle, while Thompson (2010) extends the anomaly to ECOs. Our evidence on the significant relationship between price update and initial return and the returns of the parent firm and the market intensify the debate. Moreover, as observed by Lowry and Schwert (2004), despite intense research on IPO pricing, there is relatively limited evidence regarding the determinants of price update. Hopefully, our analyses will inspire further research on Lowry and Schwert's (2004) conclusion that the initial price range is not an unbiased estimator of the offer price. ▪

Footnotes
  • 1

    Underwriters allocate shares to investors in direct proportion to their expression of interest during the book-building period. Since issues are generally oversubscribed, investors receive less than the allocation originally indicated by them. This process is often referred to as rationing, which Cornelli and Goldreich (2001) define as the number of shares allocated to a bidder divided by the number of shares he requests. An alternative normalized measure is the ratio of the percentage allocation to the percentage bid. As Cornelli and Goldreich (2003) observe, rationing allows investment bankers to compensate informed bidders through more favorable allocations.

  • 2

    Loughran and Ritter (2002) offer two potential explanations as to why underwriters prefer compensation through underpricing to charging a higher gross spread. First, a lower offer price makes the issue more marketable. Additionally, allocating underpriced shares is a quid pro quo to favored customers who return the favor by channeling more trading commissions to the underwriter.

  • 3

    Hanley (1993) finds that IPOs with offer prices above the maximum of the filing price range quoted in the preliminary prospectus have an average initial return of 21%. Issues with offer prices within the anticipated range have an average initial return of 10% and offerings at prices below the lowest anticipated price have an average initial return of 0.6%. Bradley and Jordan (2002) extend Hanley's (1993) study by examining filing range adjustments and confirm that IPOs where the file ranges are adjusted upward are more underpriced, regardless of where the offer price stands relative to the final price range. Ljungqvist and Wilhelm (2002) report that institutions that reveal more valuable information during the registration period are rewarded with larger allocations when this information is positive.

  • 4

    Loughran and Ritter (2002) measure loss due to underpricing as the number of shares sold times the difference between the opening-day closing price and the offer price, while the perceived gain in wealth is measured as the number of shares retained times the difference between the opening-day closing price and the midpoint of the original offering price range.

  • 5

    The behavioral proxy is defined as the relative size of the chief executive officer's wealth loss due to underpricing of shares sold and his (perceived) wealth gain due to the revaluation of his retained shares.

  • 6

    Further, Lowry and Schwert (2004) demonstrate that for IPOs, although the correlation between prior market returns and underpricing is statistically significant, it is not economically large enough to earn arbitrage profits. Indeed, the authors assert that IPO pricing is almost efficient with respect to public information.

  • 7

    Our objective is to analyze the IPO pricing process in the context of the unique characteristics of ECOs, especially the impact of returns of the parent firm, which holds a controlling interest in the subsidiary, on price update and initial return of the carved-out IPO. Lowry and Schwert (2004) follow a similar approach.

  • 8

    Loughran and Ritter (2004) attribute issuing firms’ acquiescence to greater underpricing in more recent years to two reasons. First, issuing firms preferred larger underwriters with more reputed analysts. Additionally, underwriters set up personal accounts for issuing firm managers. Ljungqvist and Wilhelm (2003) note that while prestigious underwriters traditionally avoided high-risk issuers, in recent years, high-tech and Internet issuers had more to gain from superior information production and the certification provided by ranked underwriters.

  • 9

    Indeed, the retention of IPO shares is highest in 2006 and 2007. However, sample sizes in these years are too small to draw any general conclusions.

  • 10

    To test for heteroskedasticity, we use the Breusch-Pagan (1979) test. To correct for heteroskedasticity, we use heteroskedasticity-consistent standard errors following White (1980).

  • 11

    Note that we also control for Dfocus, an indicator variable that is equal to one when the IPO and the parent firm have different two-digit SIC codes, and zero otherwise. We document a positive, but nonsignificant effect of Dfocus on price update, as well as initial return. Our results are similar when Dfocus is included in the models. Therefore, we do not report models including Dfocus for brevity.

  • 12

    The VIF measures the impact of collinearity among the variables in a regression model. The VIF is equal to 1/Tolerance, and it is always greater than or equal to one. On the basis of the VIF test, there is no multicollinearity issue in our models as all VIF values are less than two. Significant multicollinearity issues arise when VIF values are > 5.

  • 13

    This result suggests that unlike venture capitalists, the parent firm may not be expected to provide much certification and monitoring that would reduce uncertainty and underpricing. The alternative view that may have bearing on the result is the argument by Frank and Harden (2001) that maintaining control of a subsidiary in an industry related to the parent firm's line of business could lead to tangible or intangible operating synergies, including the joint development of shared technologies or production processes and joint procurement or production systems. However, the authors fail to find support for this hypothesis. It is worth noting that over 60% of our sample involves parent subsidiaries in dissimilar industries. Further exploration of this issue is worthy of future research.

  • 14

    We have estimated all the relevant models with only the 154 cases that were announced before the pricing day. Our results are not qualitatively changed.

References

  1. Top of page
  2. Abstract
  3. I. Pricing of ECOs—Hypotheses
  4. II. Sample and Data
  5. III. Empirical Analysis: Univariate Results
  6. IV. Multiple Regression Analysis
  7. V. Discussion and Conclusions
  8. References