Parent Firm Characteristics and the Abnormal Return of Equity Carve-outs

Transcription

1 Parent Firm Characteristics and the Abnormal Return of Equity Carve-outs Feng Huang ANR: MSc. Finance Supervisor: Fabio Braggion Second reader: Lieven Baele

2 Parent firm characteristics and the abnormal return of equity carve-outs Abstract This study investigates the impact that ex-ante parent firm characteristics have on abnormal returns around the announcement and on initial returns of carved out subsidiaries for a sample of 104 carve-outs and their parent during the period I examine the performance, leverage and corporate governance effect on abnormal returns during 3-day event window and on initial day returns of subsidiaries. The results show that parents with poor performance are likely to have higher abnormal returns during the event window, and the carved out subsidiaries are likely to obtain higher returns on the first trading day. However, the performance effect is insignificant. In addition, highly leveraged parent firms tend to have higher abnormal returns and the subsidiaries are likely to have higher initial day returns. Furthermore, this study suggests a negative relationship between the pre-event ownership concentration and initial day returns, but the governance effect on abnormal returns is not clear. Keyword: Equity carve-out, IPO underpricing, performance, leverage, corporate governance 1

3 1. Introduction Equity carve-out is a type of corporate reorganization, in which a company creates a new subsidiary and sells part of the subsidiary through an initial public offering (IPO), while retaining management control. Typically, up to 20% of the subsidiaries shares are offered to public (80% retention by the parents) if parent companies desire to conduct complete divestitures via a tax-free spin-off (Internal Revenue Code, 26 US Code Sec. 355). The transaction creates two separate legal entities, the parent firms and carved out subsidiaries each with their own boards, management teams, and financials. In this study, I follow Schipper and Smith (1986) and Vijh (1999, 2002) and define equity carve-out as a partial IPO of a subsidiary by a parent firm. In viewing equity carve-outs as IPOs, I borrow from the work of Thomson (2010) who proposes the implication that models with public information lessen the need for models that rely on information asymmetry, and investors can invest in stocks based on publicly available information. With this implication, I investigate the impact that ex-ante parent firm characteristics have on abnormal returns around the announcement and on initial returns of carved out subsidiaries for a sample of 104 carve-outs and their parent during the period I examine three hypotheses for the market reaction. First, firms with better performance before equity carve-outs tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. This hypothesis is a further step from Lang, Poulsen and Stulz (1995) and Allen and McConnell (1998). Second, firms with low leverage before equity carve-outs tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. This hypothesis is obtained from James and Wier (1990), Su (2004) and Barry and Mihov (2006), and also a further step from Lang, Poulsen and Stulz (1995) and Allen and McConnell (1998). Third, firms with better corporate governance before equity carve-outs tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. This hypothesis is obtained from Elsas and Loffler (2007) and Chahine and Zeidan (2012). 2

4 With a sample of 104 equity carve-outs during the period between 1990 and 2009, I obtain an average initial day return of 16.64% (t-statistic = ) and an average 3-day cumulative abnormal return of 1.79% (t-statistic = ). The abnormal return is similar to that in Vijh (2002), which is 1.93% (t-statistic = 5.64). Empirical analysis is then used to test the impact that parent characteristics have on initial returns and on abnormal returns. In our empirical work, I find little support for the first hypothesis. The signs in the regressions are negative, which are opposite from the hypothesis. It suggests that parents with poor performance not only tend to conduct equity carve-outs (Lang, Poulsen and Stulz (1995) and Allen and McConnell (1998)), they are also likely to have higher abnormal returns during the event window, and the carved out subsidiaries are likely to obtain higher returns on the first trading day. However, the performance effect is not significant. One possible reason for the negative relationship may be the unique characteristics of equity carve-outs, that they can be regarded as good news to firms. Thus carve-outs will bring a greater shock to poor performers than good performers. This study shows some evidence on the second hypothesis, but also in an opposite direction. Although the leverage effect is not significant on initial day returns, it is economically significant on 3-day cumulative abnormal returns when coverage ratio is used. It is suggested that highly leveraged parent firms tend to have higher abnormal returns and the subsidiaries are likely to have higher initial day returns. This is also a further result of Lang, Poulsen and Stulz (1995) and Allen and McConnell (1998) that financially constrained firms are more likely to conduct equity carve-outs. Moreover, the further check of the leverage effect shows that the short-term leverage tends to be the dominant factor that influences the abnormal returns and initial day returns. While on the other hand, some of effect is likely to be offset by the long-term leverage. Furthermore, this study suggests a negative relationship between the pre-event ownership concentration and initial day returns, which provides some support for the 3

5 third hypothesis. This is similar to the results of Elsas and Loffler (2007) in Germany and Chahine and Zeidan (2012). Still this result is not significant. However, the governance effect on abnormal returns seems to be different from the hypothesis. The share ownership of management tends to have a positive relationship with 3-day cumulative abnormal returns, which is economically significant. Two possible explanations are proposed. One is that the amount of the sample is limited. The other is that DRT may be a poor proxy for corporate governance. What is interesting is that consistent with Hogan and Olson (2004), this study also suggests that price revisions are significantly related to the initial day returns. The subsidiaries with higher price revisions are also likely to have higher returns on the first trading day. It is also suggested that the industry of carved out subsidiaries is likely to be an important factor to influence the abnormal returns of parents. The rest of the paper is organized as follows: the next section provides a literature review of equity carve-outs and presents the corresponding hypotheses related to the parent firm characteristics, initial day returns and abnormal returns. Section 3 provides the data and methodology of the paper. The variable definition and a summary description of these variables are also in this section. Section 4 illustrates empirical results and discusses them. Section 5 describes the robustness check of the results. Section six presents the main conclusions of this paper. 2. Current state of literatures 2.1 Equity carve-outs Equity carve-out is a type of corporate reorganization, in which a company creates a new subsidiary and sells part of the subsidiary through an initial public offering (IPO), while retaining management control. Typically, up to 20% of the subsidiaries shares are offered to public (80% retention by the parents) if parent companies desire to conduct complete divestitures via a tax-free spin-off (Internal Revenue Code, 26 US Code Sec. 355). The transaction creates two separate legal entities, the parent firms 4

6 and carved out subsidiaries each with their own boards, management teams, and financials. In this study, I follow Schipper and Smith (1986) and Vijh (1999, 2002) and define equity carve-out as a partial IPO of a subsidiary by a parent firm. A considerable part of prior research on equity carve-outs focuses on the performance of the parent company s stock around the time the carve-out is announced, and positive abnormal performance is recorded around equity carve-outs. With a sample of 76 equity carve-out announcements between 1965 and 1983, Schipper and Smith (1986) note that an equity carve-out is expected to have a positive share price effect due to four characteristics. These include: (1) the separation of subsidiary and parent assets for external equity financing, (2) the initiation of public trading of subsidiary common stock, (3) the restructuring of asset management and incentive contracts and (4) the creation of a publicly held minority interest. Allen and McConnell (1998) perform a study over 186 equity carve-outs over the period They propose the managerial discretion hypothesis of equity carve-outs. Using a sample of 628 carve-outs during , Vijh (1999) also find that subsidiary firms do not underperform the benchmarks even in the long run. Nanda (1991), Frank and Harden (2001), Hogan and Olson (2006), Thompson and Apilado (2009), and Semadeni and Cannella Jr. (2011) also find similar results. 2.2 Abnormal returns around announcement Given the unique performance of equity carve-outs, a lot of prior literature tries to explain it and examine the variables that may have an impact on the abnormal return around the announcement. Nanda (1991) shows that firms that resort to an equity carve-out will be firms that, on average, are being undervalued by the market. Allen and McConnell (1998) propose that firms engaging in equity carve-outs generally should have high leverage and poor operating performance. They find that the use of funds is the dominant factor of the effect in announcement period returns. Using a sample of 628 carve-outs during 5

7 , Vijh (1999) propose that the superior performance of subsidiary stocks arises because the subsidiary and parent firms can focus on fewer business segments after carveout. In a study of a sample of 336 carve-outs that occurred in the period , Vijh (2002) shows that the announcement-period returns increase with the ratio of subsidiary to non-subsidiary assets. Elsas and Loffler (2007) analyze value effects of changes in the governance structure of German firms due to equity carve-outs and find that a higher degree of pre-event ownership concentration leads to lower abnormal returns. 2.3 IPO underpricing The term IPO underpricing refers to the difference in the closing offer price of a security (through IPO), and the closing price of the security on first trading day, in case if listing price is higher. Ibbotson (1975) and others have presented evidence that initial public offerings (IPOs) of firms stocks are underpriced. Allen and Faulhaber (1989) suggest that underpricing of new issues occurs at certain times in particular industries. Loughran and Ritter (2004) note that in the 1980s, the average first-day return on IPOs was 7%. The average first-day return doubled to almost 15% during , before jumping to 65% during the internet bubble year of and then reverting to 12% during Ruslim, Marciano and Wijaya (2010) also find that IPOs are significantly underpriced at the first day of trading. The level of underpricing is also influenced by the firm s background, past performance, existing shareholding pattern, present performance, etc. Bradley and Jordan (2002) show that 35%-50% of the variation in IPO underpricing can be predicted using public information known before the offer date. They focus primarily on four variables: share overhang, file range amendment, venture capital backing, and previous issue underpricing. Ranjan and Madhusoodanan (2004) develop a model describing the IPO process in the presence of asymmetric information. This model suggests that IPO underpricing is more severe in the case of smaller issue sizes. 6

8 Ruslim, Marciano and Wijaya (2010) find that IPOs are significantly underpriced at the first day of trading. Financial institution sector s IPOs are less underpriced than non-financial institutions sectors. In addition, factors, such as type of business entities and trade price volatility in the stock market, also affect underpricing. Leverage, size, underwriter, shareholder dispersion and other variables are also considered to have a significant impact on the underpricing (Carter and Manaster (1990), James and Wier (1990), How and Low (1993), Carter, Dark and Singh (1998), Dor (2003), Su (2004), Ranjan and Madhusoodanan (2004), Barry and Mihov (2006), Pukthuanthong and Walker (2008)). 2.4 Equity carve-outs underpricing Equity carve-outs are partial IPOs in which parents sell part of their ownership in wholly owned subsidiaries. Therefore, the initial day returns of equity carve outs may exhibit similar pattern as normal IPOs. Actually, Vijh (2002) reports that pure carve-outs have lower subsidiary initial period returns than IPOs. With a sample of 458 equity carve-outs from 1990 to 1998, Hogan and Olson (2004) find the pricing of equity carve-outs to be positively related to the overallotment option, gross spread, log of market to sales ratio, use of prestigious underwriters, amount of stock dilution or share overhang, inclusion in the technology sector, and price revisions before the offering. In another study, for a sample of 219 carve-outs during the period 1991 to 2000, Hogan and Olson (2006) examine the effect the characteristics of equity carve-outs have on the initial returns. After analyzing a sample of 271 equity carve-outs in , Thompson (2010) find that 11 35% of the variation in ex-ante public information can predict equity carve-out initial returns. The information includes investment banker reputation, percentage to be retained by parents, option index volatility, market volatility and movement due to partial price adjustment. In another study of 414 carve-outs from 1988 to 2006, Thompson (2013) examine the impact of ex-ante variables on the three-year period returns of equity 7

9 carve-outs. Similar with the previous research, he find that the variables including overhang, the proportion of the offering proceeds reinvested by the company, investment banker reputation, market volatility, and partial price adjustment have a significant effect on the long-run return of equity carve-outs. 2.5 Parent-firm characteristics and hypothesis development Despite the numerous literatures that studied the ex-ante variables that influence the abnormal returns and carve-out underpricing, few literatures focused on the influence of the characteristics of parent firms on the abnormal returns or initial day returns. Hence, it is not exactly known how much and how parent firm characteristics affect the abnormal returns and initial day returns. With the presumption that before equity carve-outs, investors can only value the carved-out subsidiaries based on available information of the parent firms, it is probable that the abnormal returns and the level of underpricing can be partly predicted by the parent-firm characteristics. This paper is going to find out the relation between the ex-ante variables of parent firms and the short-term stock performance around equity carve-outs. The presumption I adopt is that investors will value both the parent firms and carved-out subsidiaries based on available public information of parent firms. And I choose three series of characteristics of the parent firms, which include performance, leverage and corporate governance. I will examine the relations between the returns and these characteristics while controlling for some other variables. i. Performance Lang, Poulsen and Stulz (1995) find that parent firms perform poorly before the asset sales. They propose the financing hypothesis to explain the abnormal returns around the announcement of asset sales. They argue that asset sales may provide a source of funds that managers are preferable to capital markets. They use return on assets (Net income/total assets), ratio of operating income to total assets, cumulative net of 8

10 market returns, Tobin s q and managerial ownership as a fraction of total equity as performance characteristics. With the payout sample of 40 sales and reinvest sample of 53 sales, they find that firms selling assets typically are poor performers. And according to managerial discretion hypothesis, Allen and McConnell (1998) predict that parent firms that undertake equity carve-outs are likely to have poor performance recently. They use the operating profit margin (EBDIT/Sales), return on total assets (EBDIT/Total assets), market to book ratio of equity and pre-carve-out HPER to measure firm performance. With a sample of 188 equity carve-outs from 1978 to 1993, they show that prior to initiating an equity carve-out, parent firms did perform poorly. Also, they claim that parent firms prior performance gives rise to the positive abnormal return of equity carve-outs. Therefore, it is likely that the performance of parent firms should have a positive relation with the abnormal returns of equity carve-outs. However, there is no previous literature on the relationship between parent firm performance and the subsidiary initial day returns. With the presumption that investors will invest according to the available information of parent firms, it is likely that the public will run to buy the stocks with better performed parent firms, and thus push up the prices and returns. Hence, it can be proposed that performance of parent firms should have a positive relation with the initial day returns of carved out subsidiaries. The first hypothesis I propose is as follows: H1: Firms with better performance before equity carve-outs tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. ii. Leverage According to the financing hypothesis, Lang, Poulsen and Stulz (1995) find that firms selling assets are usually constrained to the capital markets due to their high leverage. They use coverage ratio, ratio of short-term liabilities to total assets, ratio of short-term debt to total assets, ratio of long-term debt to total assets and leverage ratio 9

11 as leverage measurements. While according to managerial discretion hypothesis, Allen and McConnell (1998) predict that parent firms that undertake equity carve-outs are likely to be highly leveraged. They also claim that managers undertake equity carve-outs only when the firm is capital constrained. Thus, parent firms are likely to be highly leveraged. Also, they claim that parent firms prior performance, including financing condition, gives rise to the positive abnormal return of equity carve-outs. Similarly with Lang, Poulsen and Stulz (1995), they also use coverage ratio, ratio of long-term debt to total assets and leverage ratio to measure leverage. Therefore, it is likely that the leverage of parent firms should have a negative relation with the abnormal returns of equity carve-outs. In addition, James and Wier (1990) present and test a model that illustrates how the existence of a borrowing relationship can reduce the ex ante uncertainty about the value of the issuing firm s equity in the secondary market and thereby reduce underpricing. The model predicts that firms with inside debt will experience less severe underpricing when they go public. Furthermore, Barry and Mihov (2006) confirm the theories of James and Wier (1990) and document a strong negative relation between debt financing and initial returns. Since equity carve-out is partial IPO, it can be proposed that the leverage of parent firms should have a negative relation with the initial day returns of carved out subsidiaries. The second hypothesis I propose is as follows: H2: Firms with low leverage tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. iii. Corporate governance According to Gompers, Ishii and Metrick (2003) and Bauer, Gunster and Otten (2004), well governed firms have higher equity returns. Inspired by this, it can be inferred that investors may be willing to pay a premium for the subsidiary stocks with well governed parent firms. In addition, by studying a sample of 54 equity carve-outs from 1984 to 2004 in Germany, Elsas and Loffler (2007) find that a higher degree of 10

12 pre-event ownership concentration, which can be a proxy for worse corporate governance, leads to lower abnormal return. Furthermore, by using the Gompers, Ishii, and Metrick corporate governance index on a sample of 158 parent firms, Chahine and Zeidan (2012) demonstrate that firms with superior governance have a better short term price reaction to carve-out announcements. Therefore, it is likely that the corporate governance of parent firms should have a positive relation with the abnormal returns of equity carve-outs. Moreover, with the presumption that investors will invest according to the available information of parent firms, the subsidiaries with better governed parents are likely to be regarded to be well governed in the future. Therefore, it is likely that the public will be willing to pay a premium for the stocks of these subsidiaries, and thus push up the prices and result in higher initial day returns. Nevertheless, Loughran and Ritter (2004) state that the increased size of friends and family share allocation, which can be regarded as poor corporate governance, may make issuing firm decision-makers less motivated to bargain for a higher offer price, and thus lead to high underpricing of IPOs. It is not known which effect will be dominant in the initial day returns of equity carve-outs. In this study, I adopt the former argument and propose that there exists a positive relationship between the governance of parent firms and initial day returns of subsidiaries. The third hypothesis I propose is as follows: H3: Firms with better corporate governance before equity carve-outs tend to have higher abnormal returns and their subsidiaries tend to have higher initial day returns. 3. Data and methodology 3.1 Sample data Several sources are used to identify the sample of equity carve-outs. A list of subsidiary IPOs in US from 1990 to 2009 was obtained from the Thomson One 11

13 database. According to the difference between equity carve-outs and spin-offs (Schipper and Smith (1986)), I choose the subsidiary IPOs from the category of spin-offs. This process identifies 414 candidates. After adjusting for divestiture IPOs, there are 298 carve-outs left. Furthermore, to enter the final sample of equity carve-outs, a candidate must satisfy the following 2 criteria: (1) Both the subsidiary and parent firms must be listed in Nasdaq, NYSE or Amex (142 carve-outs eliminated), and (2) data for the parent firms and subsidiaries must have been available on the Center for Research in Security Prices (CRSP) and COMPUSTAT (52 carve-outs eliminated). This provides a sample of 104 carve-outs. The announcement dates were obtained from the Factiva. The COMPUSTAT provides the accounting data of parent firms, the CRSP provides the stock market data, and the SEC filings provide the shareholders distribution information. 3.2 Methodology In order to assess the market reaction to equity carve-outs, both Initial Day Returns (IDR) of subsidiary stocks and Cumulative Abnormal Returns (CAR) measures are used. Consistent with Schipper and Smith (1986) and Thompson (2010), the IDR is defined as the closing price on the first trading day to the offering price. I adopt the market model of Vijh (2002) to calculate the abnormal returns around day 0. Day 0 denotes the day when the information first hit the market, which is the earlier of the announcement date and filing date. In this sample, day 0 represents the announcement date in 22 candidates, the filing date in 29 candidates, and a simultaneous date in 53 candidates. According to the results of Vijh (2002), the event window for abnormal returns is from day -1 to day 1. The abnormal returns (AR) are the difference between the stock returns (R) and the 12

14 expected normal returns (NR) of the sample firms. AR it = R it NR it The market model below is used for the expected normal returns of each sample firm: NR it = a i + b i R mt Where NR it = return on stock i, day t from CRSP, a i, b i = estimated market model parameters for sample firm i, R mt = returns of market index on the stock market where sample firm i is in, The parameters a i and b i are estimated over a 250-day period ending 250 days before day 0. The market returns used here are measured by the returns on the CRSP value-weighted portfolio of all stocks. The calculated 3-day abnormal return from day -1 to day 1 is 1.79% (t-statistic = ), which is consistent with the results of Vijh(2002). Table 1 represents the annual distribution of the sample. It is clearly shown that most of the carve-outs were taken in the 1990s. It shows that the stocks of the subsidiaries can earn 16.64% in the first trading day. And during the 3-day event window, the stocks of the parent firms have a cumulative abnormal return of 1.79%. The ordinary least square regression is used to test the hypotheses. The models I employ are as follows: IDR = β 0 + β 1 Explanatory Variables + β 2 Controlled Variables CAR3 = β 0 + β 1 Explanatory Variables + β 2 Controlled Variables 13

15 Table 1: Annual distribution of carve-outs This table reports the annual distribution of equity carve-outs during the period 1990 through Data are obtained from Thomson One database and CRSP. The initial day returns are defined as the closing price on the first trading day to the offering price. The 3-day cumulative abnormal returns are obtained from the model of Vijh (2002). Carve out year Number of carve-outs Average Initial Day Returns Average 3-day Cumulative Abnormal Returns % 3.59% % 3.18% % -0.75% % -0.58% % 3.71% % 0.40% % 3.86% % 3.76% % 0.74% % 1.46% % 3.05% % -0.63% % 3.63% % 0.07% % 0.09% % 3.56% % 0.32% % -1.15% % 0.14% Total % 1.79% Due to the time difference between IDR and CAR3, the controlled variables are different in the two regressions. The controlled variables in the first regression include USE, INDUSTRY, UNDERWRITER, REVISION, RELATIVE, SIC, OVERALLOTMENT and GROSSSPPREAD. REVISION is absent in the second regression. The definitions of both explanatory and controlled variables are in the following part. 14

16 3.3 Variable definition The definitions of explanatory variables are as follows: I adopt several different variables to study the performance effect according to Lang, Poulsen and Stulz (1995), and Allen and McConnell (1998). The variable profit margin (PM) is defined as the ratio of net income to sales. Operating profit margin (OPM) is defined as the ratio of EBIT to sales. Return on asset (ROA) is defined as the ratio of net income to total assets. Return on equity (ROE) is defined as the ratio of net income to equity. PM and ROA are used to test H1, while OPM and ROE are used to check the robustness. And according to H1, the parameters of these variables are likely to be positive. Similarly, several different variables are used to study the leverage effect according to Lang, Poulsen and Stulz (1995), and Allen and McConnell (1998). The variable coverage ratio (COVERAGE) is defined as the ratio of EBIT to total interest payments. Leverage ratio (LEVERAGE) is defined as the ratio of total debt to total assets. Long-term leverage (LEVERAGEL) is defined as the ratio of long-term debt to total assets. Short-term leverage (LEVERAGES) is defined as the ratio of short-term debt to total assets. LEVERAGEL and LEVERAGES are also used to check the robustness. And according to H2, the parameter of COVERAGE is likely to be negative, while the parameters of others are likely to be positive. Similar to Elsas and Loffler (2007), I use the concentration of share holdings to proxy for the corporate governance. A higher degree of concentration implies poor corporate governance, and thus is likely to decrease the abnormal returns. I employed three methods to measure the concentration. The first is the Herfindahl-Hirschman Index (HHI), the other is the Concentration Ratio (CR) and the last is the percent of shares held by the management (DRT). Therefore, the variable HHI3 is defined as the sum of the squares of the share holdings of the largest 3 shareholders. CR3 is defined as the sum of the share holdings of the largest 3 shareholders. According to H3, the parameters of CR3 and DRT are likely to be negative. HHI3 is used for robustness 15

17 check. Table 2 summarizes the characteristics of the explanatory variables for CAR3. It shows that, ROA has a mean of and a median of , indicating a left fat tail. It is consistent with the result of Lang, Poulsen and Stulz (1995) and Allen and McConnell (1998) that parent firms tend to have poor performance before equity carve-outs. The large value of standard deviation of COVERAGE ( ) indicates that the interest coverage ratio varies a lot among different parents. The explanatory variables for IDR exhibit similar characteristics. I don t put in the thesis for the consideration of conciseness. Table 2: Summary of explanatory variables for CAR3 This table presents summary statistics for the explanatory variables for CAR3 used in this paper. PM is the ratio of net income to sales. ROA is the ratio of net income to total assets. COVERAGE is the ratio of EBIT to total interest payments. LEVERAGE is the ratio of total debt to total assets. CR3 is the sum of the share holdings of the largest 3 shareholders. DRT is the percent of shares held by directors. Variable Mean Std. Dev. Min 25% Median 75% Max PM ROA COVERAGE LEVERAGE CR % 17.30% 0.12% 10.90% 21.01% 31.80% % DRT 10.05% 15.98% 0.01% 1.70% 4.06% 12.50% % To control the effect of other possible factors, I select several variables according to the previous study. The definitions of controlled variables are as follows: Lang, Poulsen and Stulz (1995) claim that the average stock price reaction is positive only when the proceeds are paid out. Allen and McConnell (1998) find that the use of the funds is the dominant factor to influence the abnormal return. Vijh (2002), 16

18 Thompson (2010) and Chahine and Zeidan (2012) also find that the market reaction is related to the use of proceeds. In addition, Leone, Rock and Willenborg (2007) suggest that the disclosures of usage have an impact on initial day returns of IPOs. Thus, the dummy variable USE is used to control for the effect of the usage of proceeds. Following Lang, Poulsen and Stulz (1995), Allen and McConnell (1998), Thompson (2010) and Chahine and Zeidan (2012), this variable is equal to 1 if the parent firms use all or part of the proceeds to repay debt or to pay shareholders, and 0 otherwise. It is used in both regressions. Elsas and Loffler (2007) and Vijh (2002) show that whether the industries of the parent and subsidiary are same also has an impact on the carve-out due to the existence of negative synergies. Hence, the dummy variable INDUSTRY is introduced to indicate the industry relation between subsidiaries and parent firms. It equals to 1 if the 2-digit SIC codes of parent firms and of subsidiaries are same, and equals to 0 if different. It is used in both regressions. Carter and Manaster (1990), Carter, Dark and Singh (1998), Prezas, Tarimcilar and Vasudevan (2000) show that the initial-day returns are lower for issues conducted by high prestige investment bankers. Hogan and Olson (2004) also find that the abnormal returns are related with the use of underwriters. Thus, the variable UNDERWRITER is used to control for the underwriter effect on the abnormal returns. Following Prezas, Tarimcilar and Vasudevan (2000), I use the updated version of Carter and Manaster (1990) from Loughran and Ritter (2004) to proxy for the investment banker reputation. Because this ranking is from 1980 to 2011, I use the corresponding ranking for different periods. It is used in both regressions. Thompson (2010) suggest that market-adjusted returns can be partly predicted by and partial price adjustment. Hogan and Olson (2004) find that price revisions are related to the pricing of equity carve-outs. Due to the similarities of the two measures, I adopt the variable of price revisions as Hogan and Olson (2004). The variable REVISION is defined as the difference between the offer price and the original middle of file price range divided by the original middle of file price range in percent form. It is only used 17

19 in the first regression. Vijh (2002) and Hogan and Olson (2006) show that the relative size of subsidiary has a positive relation with the abnormal return during the announcement period. Ranjan and Madhusoodanan (2004) also suggest that IPO underpricing is related to issue size. Therefore, the relative size RELATIVE is defined following Vijh (2002) and Hogan and Olson (2006). It equals the offering size divided by the market value of the parent firm in percent form. It is used in both regressions. Hogan and Olson (2004) claim that the pricing of equity carve-outs is positively related with the inclusion of technology sector. Perotti and Rossetto (2007) argue that different industries should treat equity carve-outs differently. Pukthuanthong and Walker (2008) find different leverage effect for firms in low- and high-tech industries. Ruslim, Marciano and Wijaya (2010) also find different underpricing patterns for firms in different sectors. Thus, an industry dummy variable is introduced to control for the industry effect for the equity carve-outs. The variable SIC is defined as the 2-digit SIC codes of subsidiaries. It is used in both regressions. Overallotment option, gross spread is also found related with abnormal return by Hogan and Olson (2004, 2006). The variable OVERALLOTMENT is defined following Hogan and Olson (2004, 2006). It is the percentage of overallotment shares sold as the percent of the total amount of shares. The variable GROSSSPREAD is also defined following Hogan and Olson (2004, 2006) and it is the total expenses as the percent of total proceeds. They are also used in both regressions. Therefore, the two regressions I use are as follows: IDR = β 0 + β 1 Explanatory Variables + β 2 USE + β 3 INDUSTRY + β 4 UNDERWRITER + β 5 REVISION + β 6 RELATIVE + β 7 SIC + β 8 OVERALLOTMENT + β 9 GROSSSPPREAD CAR3 = β 0 + β 1 Explanatory Variables + β 2 USE + β 3 INDUSTRY + β 4 UNDERWRITER + β 5 RELATIVE + β 6 SIC + β 7 OVERALLOTMENT + β 8 GROSSSPPREAD 18

20 4. Empirical analysis 4.1 Descriptive statistics Table 3 summarizes the characteristics of initial day returns and 3-day cumulative abnormal returns. It can be clearly seen that a typical IDR of 16.64% (t-statistic = ) on average. The IDR is positively fat tailed with a median of only 5.37%. Similarly, the CAR3 also has a positive fat tail with a mean of 1.79% (t-statistic = ) and a median of 0.99%. The results are consistent with Schipper and Smith (1986), Allen and McConnell (1998), Vijh (1999), Nanda (1991), Frank and Harden (2001), Hogan and Olson (2006), Thompson and Apilado (2009), and Semadeni and Cannella Jr. (2011). Therefore, it can be reasonably inferred that equity carve-outs tend to be good news for firms. Table 3: Summary of dependent variables This table presents summary statistics for the dependent variables used in this paper. IDR is the initial day return, and is defined as the closing price on the first trading day to the offering price. CAR3 is the 3-day cumulative abnormal return, and is obtained from the model of Vijh (2002). Variable Mean Std. Dev. Min 25% Median 75% Max IDR 16.64% 34.49% % 0.00% 5.37% 17.09% % CAR3 1.79% 6.64% % -1.17% 0.99% 4.66% 41.37% For every single variable, I sort the sample and divided it into three subsamples to calculate the mean and median IDRs and CAR3s. Table 4 presents the descriptive statistics for the subsamples. It shows that the parent firm characteristics, except for the COVERAGE, have no significant impact on the IDRs. In the subsample with lowest coverage ratio, the mean IDR is 7.89% and the median is 2.94%. While in the 19

21 This table presents univariate sorting results for the IDR and CAR3. The t-statistics indicate the statistical significance of the difference between the means of the lowest and highest groups. PM is the ratio of net income to sales. ROA is the ratio of net income to total assets. COVERAGE is the ratio of EBIT to total interest payments. LEVERAGE is the ratio of total debt to total assets. CR3 is the sum of the share holdings of the largest 3 shareholders. DRT is the percent 20 Table 4: Descriptive results of shares held by directors. ***, **, and * indicate significance at the 1%, 5% and 10% levels, respectively. Variables PM ROA Distribution IDR CAR3 Mean Median Mean Median Lowest 20.17% 4.69% 3.50% 3.24% Middle 14.27% 2.09% 1.32% 0.05% Highest 15.71% 11.10% 0.28% 0.08% Lowest 20.53% 4.69% 3.09% 1.55% Middle 11.81% 3.70% 0.43% 0.12% Highest 17.81% 8.24% 1.59% 0.54% t-statistics for differences between the IDRs in lowest and highest groups t-statistics for differences between the CAR3s in lowest and highest groups Observations

22 21 Table 4 (Continued) Lowest 7.89% 2.94% 1.53% 3.61% COVERAGE Middle 17.49% 2.00% 1.94% 0.07% LEVERAGE CR3 DRT Highest 23.84% 11.76% 1.70% 0.32% Lowest 17.49% 9.14% 1.78% 1.05% Middle 18.07% 5.73% 0.62% 0.07% Highest 14.30% 2.50% 3.01% 1.94% Lowest 14.64% 4.40% 1.78% 0.11% Middle 25.77% 7.14% 1.46% 0.89% Highest 13.07% 10.00% 3.20% 3.25% Lowest 18.47% 7.78% 1.31% 0.86% Middle 21.99% 6.31% 0.20% -0.26% Highest 13.26% 5.26% 4.61% 3.42%

23 subsample with highest coverage ratio, the mean IDR is 23.84% and the median is 11.76%. It seems that the subsidiaries that have parents with higher coverage ratios tend to have higher initial day returns, which is consistent with the hypothesis. As for the CAR3s, it also seems that parent firm characteristics, except for the PM and DRT, have no impact on them. Moreover, when the CAR3s are sorted into three subsamples according to the PM, the CAR3s seem to change in the opposite direction. In the subsample with lowest profit margin, the mean CAR3 is 3.5% and the median is 3.24%. While in the subsample with highest profit margin, the mean CAR3 is 0.28% and median is 0.08%. It seems that better performed parents tend to have lower 3-day cumulative returns. In the subsample with lowest DRT, the mean CAR3 is 1.31% and the median is 0.86%. Compared with the subsample with highest DRT, it has significantly lower abnormal returns at 10% level. It is likely that parent firms with lower DRT, which seem to be better governed, would have lower abnormal returns. It is also different from the hypothesis. What also needs attention is that most of the changes are not monotonic. These findings indicate that most of the parent firm characteristics seem to have little or no impact on the abnormal returns in the event window and initial day returns of subsidiaries. 4.2 Multivariate regressions To better test the hypotheses and avoid the effect of extreme values, I winsorize both the dependent and explanatory variables at 1% level. Also, I cluster the tests at the parent firm level in all of the following regressions Test of variables on initial day returns The results of multivariate tests for the impact of parent firm characteristics on initial day returns are presented in Table 5. Model 1 and 2 investigate the performance effect on IDR with PM and ROA as parent-firm performance measurement. Model 3 and 4 22

24 This table presents the results from ordinary least square regressions of initial day returns on parent firm characteristics. Model 1 and 2 investigate the performance effect on IDR with PM and ROA as parent-firm performance measurement. Model 3 and 4 investigate the leverage effect on IDR with COVERAGE and LEVERAGE as parent-firm leverage measurement. Model 5 and 6 investigate the corporate governance effect on IDR with CR3 and DRT as parent-firm governance measurement. Model 7 and 8 conducts multivariate test with different explanatory variables simultaneously. PM is the ratio of net income to sales. ROA is the ratio of net income to total assets. COVERAGE is the ratio of EBIT to total interest payments. LEVERAGE is the ratio of total debt to total assets. CR3 is the sum of the share holdings of the largest 3 shareholders. DRT is the percent of shares held by directors. USE is equal to 1 if the parent firms use all or part of the proceeds to repay debt or to pay shareholders, and 0 otherwise. INDUSTRY is equal to 1 if the 2-digit SIC codes of parent firms and of subsidiaries are same, and equals to 0 if different. UNDERWRITER is the ranking from the updated version of Carter and Manaster (1990) from Loughran and Ritter (2004). REVISION is the difference between the offer price and the original middle of file price range divided by the original middle of file price range in percent form. RELATIVE is equal to the offering size divided by the market value of the parent firm in percent form. SIC is the 2-digit SIC codes of subsidiaries. OVERALLOTMENT is the percentage of overallotment shares sold as the percent of the total amount of shares. GROSSSPREAD is the total expenses as the percent of total proceeds. All of the dependent and independent variables are winsorized at 1% level. Standard errors are adjusted for clustering by parent firm industry. Standard errors are reported in parentheses. ***, **, and * indicate significance at the 1%, 5% and 10% levels, respectively. 23 Table 5: Multivariable analysis on initial day returns

25 Variables (0.0835) (0.0710) 24 Table 5 (Continued) PM (0.0977) ROA (0.4170) COVERAGE (0.0030) LEVERAGE (0.1871) CR (0.2393) DRT (0.1866) (0.1064) (0.2101) (0.2563) (0.1615) (0.5900) (0.0040) (0.2353) (0.2510) (0.3691) USE (0.0480) (0.0526) (0.0540) ( (0.0562) * (0.0508) INDUSTRY (0.0451) (0.0490) (0.0609) (0.0393) (0.0566) (0.0538) (0.0634) (0.0591)

26 Variables ** (0.4402) (0.2549) (0.0014) * (0.6134) (3.6560) F Statistic ** (0.3435) (0.2382) (0.0021) * (0.9015) (5.5618) Prob (F Statistic) Adj R-squared N Table 5 (Continued) UNDERWRITER (0.0146) (0.0127) (0.0124) (0.0119) (0.0137) (0.0120) (0.0175) (0.0239) REVISION *** (0.2662) *** (0.2632) *** (0.2354) *** (0.2868) ** (0.4398) ** (0.4032) RELATIVE (0.2176) (0.2204) (0.2126) (0.2162) (0.2411) (0.2569) SIC (0.0009) (0.0010) (0.0014) (0.0009) (0.0012) (0.0011) OVERALLOTMENT * (0.9552) * (0.8924) * (1.0418) (1.0191) (0.5465) * (0.5854) GROSSSPREAD (2.2300) (2.1925) (3.8804) (2.3188) (3.8764) (2.8119)

27 investigate the leverage effect on IDR with COVERAGE and LEVERAGE as parent-firm leverage measurement. Model 5 and 6 investigate the corporate governance effect on IDR with CR3 and DRT as parent-firm governance measurement. Model 7 and 8 conduct multivariate tests with different explanatory variables simultaneously. It shows that none of the tested variables are statistically significant in all of these regressions. There seems to be hardly any performance, leverage or governance effects on IDR. What is interesting is that the signs of performance measures, whichever is taken, are negative and in the opposite direction as expected in Model 1 and 2. It indicates that if a parent firm has worse performance before equity carve-outs, its subsidiary tends to have higher initial days return. This result is contradictory against the first hypothesis. One of the possible explanations for the negative signs is that with significantly positive IDR, equity carve-outs can be regarded as good news for firms. Therefore, conducting equity carve-outs brings a greater shock to firms with poor prior performance. Equity carve-outs can be regarded by investors as an indication of future performance improvement, which is more important for poor performers. Thus, subsidiaries with worse performed parents tend to have higher initial day returns. The signs of the leverage measures in Model 3 and 4 are different from the hypothesis. It indicates that, at some level, subsidiaries with highly leveraged parent firms tend to have higher initial day returns. This is inconsistent with results of James and Wier (1990) and Barry and Mihov (2006) in normal IPOs, which may be due to the similar reason of the performance effect, which is that equity carve-outs may bring a greater shock to highly leveraged firms. However, the leverage effect is not statistically significant. Recall from the initial results from descriptive statistics, COVERAGE is in an opposite direction, and the average IDR in the largest part is significantly higher than in the smallest part. This seemingly inconsistent result may be due to the limited amount of the sample. In addition, Model 5 and 6 shows that corporate governance tends to not to have 26

28 effect on initial day returns. Except for CR3 in the last model when all explanatory variables are tested, the signs of governance measures are all negative. With negative signs, subsidiaries with better governed parents are likely to have higher initial day returns. This result is consistent with the hypothesis but inconsistent with Loughran and Ritter (2004) in normal IPOs. What is remarkable is that in the controlled variables, REVISION is statistically significant in all models at least at 5% level, which is consistent with Hogan and Olson (2004). In addition, OVERALLOTMENT is also statistically significant at 10% level in most models. In conclusion, the multivariate tests on initial day returns provide some evidence for Hypothesis 3. However, the result is opposite from Hypothesis 1 and Hypothesis 2, which may be due to the unique characteristics of equity carve-outs, and that poor performers tend to suffer greater shocks from similar good news Test of variables on abnormal returns The results of multivariate tests for the impact of parent firm characteristics on 3-day cumulative abnormal returns are presented in Table 6. Model 1 and 2 investigate the performance effect on CAR3 with PM and ROA as parent-firm performance measurement. Model 3 and 4 investigate the leverage effect on CAR3 with COVERAGE and LEVERAGE as parent-firm leverage measurement. Model 5 and 6 investigate the corporate governance effect on CAR3 with CR3 and DRT as parent-firm governance measurement. Model 7 and 8 conduct multivariate tests with different explanatory variables simultaneously. The table shows that there seem to be some leverage and governance effect when COVERAGE and DRT are tested in Model 3 and Model 6, respectively. Nevertheless, when all variables are tested simultaneously in Model 7 and 8, none of the effect exists. Model 1 and 2 show that there seems to be hardly any performance effects on CAR3. 27

29 This table presents the results from ordinary least square regressions of 3-day cumulative abnormal returns on parent firm characteristics. Model 1 and 2 investigate the performance effect on CAR3 with PM and ROA as parent-firm performance measurement. Model 3 and 4 investigate the leverage effect on CAR3 with COVERAGE and LEVERAGE as parent-firm leverage measurement. Model 5 and 6 investigate the corporate governance effect on CAR3 with CR3 and DRT as parent-firm governance measurement. Model 7 and 8 conducts multivariate test with different explanatory variables simultaneously. PM is the ratio of net income to sales. ROA is the ratio of net income to total assets. COVERAGE is the ratio of EBIT to total interest payments. LEVERAGE is the ratio of total debt to total assets. CR3 is the sum of the share holdings of the largest 3 shareholders. DRT is the percent of shares held by directors. USE is equal to 1 if the parent firms use all or part of the proceeds to repay debt or to pay shareholders, and 0 otherwise. INDUSTRY is equal to 1 if the 2-digit SIC codes of parent firms and of subsidiaries are same, and equals to 0 if different. UNDERWRITER is the ranking from the updated version of Carter and Manaster (1990) from Loughran and Ritter (2004). RELATIVE is equal to the offering size divided by the market value of the parent firm in percent form. SIC is the 2-digit SIC codes of subsidiaries. OVERALLOTMENT is the percentage of overallotment shares sold as the percent of the total amount of shares. GROSSSPREAD is the total expenses as the percent of total proceeds. All of the dependent and independent variables are winsorized at 1% level. Standard errors are adjusted for clustering by parent firm industry. Standard errors are reported in parentheses. ***, **, and * indicate significance at the 1%, 5% and 10% 28 Table 6: Multivariable analysis on 3-day cumulative abnormal returns levels, respectively.