HOW INVESTORS SECURE IPO ALLOCATIONS* Sturla Fjesme Melbourne University. Roni Michaely Cornell University and the Interdisciplinary Center

Transcription

1 HOW INVESTORS SECURE IPO ALLOCATIONS* Sturla Fjesme Melbourne University Roni Michaely Cornell University and the Interdisciplinary Center Øyvind Norli BI Norwegian Business School This version: February 13, 2013 Abstract: Using data, at the investor level, on the allocations of shares in initial public offerings (IPOs), we document a strong positive relation between the amount of stock-trading commission and the number of shares an investor receives in a subsequent IPO. We find no evidence to support the idea that investment banks allocate shares to investors that are perceived to be long-term investors. Our findings are consistent with the view that investment banks are able to capture some of the profits earned by investors when participating in underpriced IPOs. JEL classification: G24; G28 Keywords: IPO allocations; Equity issue; Commission; Rent-seeking * We are grateful to Øyvind Bøhren, François Derrien, Espen Eckbo, Douglas Foster (discussant), Bruce Grundy, Michelle Lowry, Ron Masulis (discussant), Jay Ritter, Selim Topaloglu, seminar participants at the BI Norwegian Business School, Melbourne University, the 2012 Financial Research Network (FIRN) Conference, the 2012 Australian National University (ANU) Summer Conference, for valuable suggestions, "The Center for Corporate Governance Research (CCGR)" at BI Norwegian Business School for financial support, the Oslo Stock Exchange VPS for providing the data and the investment banks and companies that helped us locate the listing prospectuses. All errors are our own. Correspondence: The University of Melbourne, 198 Berkeley Street, Victoria, Australia, 3010, Telephone: , sturla.fjesme@unimelb.edu.au.

2 1. Introduction The Securities and Exchange Commission (SEC) and the National Association of Securities Dealers (NASD) have since the early 2000s investigated the initial public offering (IPO) allocation practices of several major investment banks. One concern is that IPO allocations are tied to excessively large stock-trading commissions and that such a practice would constitute illegal kickbacks from investors to investment banks. 1 Reuter (2006) points out that such kickbacks would allow the underwriter to share more of the benefits of underpriced IPOs -and, therefore, exacerbate the agency conflict that exists between the issuing firm and the lead underwriter of the IPO. This paper investigates whether or not investors that has generated large stock-trading commissions in the past receives a preferential treatment in future IPO allocations. Using data on the stock-holdings for every single investor that owned common shares that was listed or became listed on the Oslo Stock Exchange (OSE) during the period 1993 through 2007, we are able to link stock-trading commission and IPO allocation at the investor level. The main finding of the paper is a strong and robust positive relation between the level of stock-trading commission generated by an investor prior to the IPO and the number of shares the same investor receives through the IPO allocation. It can be argued that large investors that generate more commissions are likely to apply for more IPO shares. However, the economic and statistical significance of the relation between commission and allocation is robust to controlling for the market value of the investors portfolio, as well as to other investor characteristics. We conclude that investors generating large stock-trading commission receives the most IPO shares 1 See the January 9, 2003 NASD settlement release es/2003/p

3 because of the commission they generate. Other investor characteristics are of less or no importance for IPO allocations. The empirical research on the allocation practices of investment banks have been hampered by the lack of data on IPO allocations. 2 Since information about stock-trading commissions are equally hard to come by, there is little empirical research on the relation between commissions and allocations. One exception, and the paper closest to ours, is Reuter (2006) who finds a positive correlation between stock-trading commission paid by mutual funds to lead investment banks and the holdings in IPOs underwritten by the same banks. This suggests, in general, that having a business relationship with the lead underwriter increases the chance of getting shares in underpriced IPOs. In particular, it suggests that investors can "buy" allocations by channelling their trades through the brokerage arm of the lead underwriter. In other closely related papers, Nimalendran, Ritter and Zhang (2006) show that there is a positive relation between money left on the table in IPOs and trading volume in liquid shares around IPO allocation dates and Goldstein, Irvine and Puckett (2010) show that large institutional investors pay unusually high trading commissions to lead underwriters of upcoming profitable IPOs. The strength of our paper, compared to the existing literature, is that we are able to analyse actual IPO allocations at the investor level. In the main part of the paper, we study 27,880 IPO allocations in 49 different IPOs. 3 The existing literature has suggested at least three potential explanations for what determines investment banks' decision of which investors are getting shares in oversubscribed IPOs. First, Benveniste and Spindt (1989) suggest that 2 See Ritter (2003) and Jenkinson and Jones (2004) for papers that summarizes IPO allocation studies. 3 In other words, there are 27,880 unique investor-ipo combinations in our data. In robustness tests, we also study 159,712 investor-ipo pairs where the IPO allocation data might be contaminated with some post-ipo trading. 3

4 investment banks allocate IPO shares to informed investors in return for a truthful revelation of their valuation of the issuer. Second, investment banks themselves tend to argue that they are looking for long-term investors. Third, investment banks allocate shares to investors that can provide some form of kickback. The empirical literature provides mixed results in terms of understanding IPO allocations in the light of the above three potential explanations. An important contribution of our paper is that we examine and contrast all three potential explanations simultaneously. As already mentioned, our data strongly support the view that investors can secure themselves IPO allocations through large stock-trading commissions. We find some evidence against a preferential treatment of buy-and-hold investors. We find some support for the idea that institutional investors get allocation in return for revealing private information about issuing firm value. The rest of paper is organized as follows: Section 2 describes the related literature. Section 3 describes testable implications. Section 4 describes the data set. Section 5 gives the empirical results, and section 6 concludes the paper. 2. Related literature The existing literature on IPO allocations provides, in general, three main explanations towards how IPO shares are allocated; see Ritter (2003) and Jenkinson and Jones (2004). First, is the information view based on Benveniste and Spindt (1989). Under this view, investment banks allocate IPO shares to informed investors in return for valuation information about the issuer. Second, is the pitchbook view where investment banks allocate shares to investors that are 4

5 perceived to be buy-and-hold. It is argued, by the investment banks that compete for IPO contracts, that allocations will be directed away from flipping investors towards more long term investors. Finally, is the rent-seeking view where investment banks will allocate shares to investors in return for kickbacks. The existing literature have found support for each of the three views and against the information view and the pitchbook view. Most studies investigate one of the views separately. The main exception is Jenkinson and Jones (2004) who directly compares the information view to the pitchbook view. Several studies investigate the information view alone. Cornelli and Goldreich (2001) investigate the order book of 23 and 16 international IPOs and SEOs. They find that regularly participating, large bid and domestic participants are favoured in allocations. They also find that bidders that participate in both hot and cold issues are given larger allocations in hot issues. Cornelli and Goldreich (2003) investigate the order book of 37 and 26 international IPOs and SEOs. They find that bids from large, frequent bidders that include a limit price affect the issue price. It is concluded that book-building is designed to extract valuation information about the issuer from the investors. Ljungqvist and Wilhelm (2002) look at the aggregated IPO allocations between institutional and retail investors for 1,032 international IPOs. They find that institutional investors are favoured over retail investors. They find that an increased institutional allocation percentage is linked to a higher deviation from the midpoint of the book-building pricing range to the final offer price. It is concluded that underwriters use institutional bids to set the IPO offer price. Binay, Gatchev and Pirinsky (2007) investigate 4,668 U.S. IPOs and find that underwriters favour institutions they have previously worked with. A relationship with the underwriter is more important in IPOs with strong demand, IPOs of less liquid firms and IPOs by less famous underwriters. It is argued that favouring regular investors is done to price IPOs more correctly. 5

6 Regular investors have incentives to report their true value in the book-building so that they will be favoured in future IPOs. Bubna and Prabhala (2007) investigate 137 Indian IPO allocations. They find that book-building and discretion in allocation enhances pre-market price discovery. All these papers are consistent with the information view of IPO allocations. The two main papers that investigate the pitchbook view of IPO allocations are Jenkinson and Jones (2004) and Aggarwal (2003). Jenkinson and Jones (2004) study 27 European IPO order books to compare the pitchbook view to the information view. They find that there is limited information gathering in the book-building procedure. This is inconsistent with the information view. Jenkinson and Jones (2004) do, however, find evidence in favour of the pitchbook view and concludes that IPO allocations are made to buy-and-hold investors. Aggarwal (2003) investigate the pitchbook view by looking at flipping by institutional and retail investors in 193 U.S. IPOs. She finds that institutional investors flip more than retail investors. This is taken as evidence against the pitchbook view. Under this view institutions are allocated more IPO shares because they are more likely to be buy-and-hold than retail investors. There are four types of IPO rent-seeking that have led to investigations (and settlements) between the SEC or NASD and different investment banks; see Liu and Ritter (2010). IPO allocations can be tied to future corporate business (IPO spinning), after-listing purchases of IPO shares (IPO laddering) or stock-trading commissions. Issuing companies can also agree to heavy underpricing in return for after-listing coverage from a star analysts provided by the investment bank (analyst conflict of interest). The heavily underpriced shares are then allocated to clients that generate high stock-trading commissions. In this way the allocating investment bank is able to recapture some of the underpricing. All four of the rent-seeking allocation practices, investigated by the SEC, have recently been looked at in different empirical studies. Liu and 6

7 Ritter (2010) investigate IPO spinning, Griffin, Harris and Topaloglu (2007) and Fjesme (2011) investigate IPO laddering, Cliff and Denis (2004) and Liu and Ritter (2011) investigate analyst conflict of interest and Reuter (2006), Nimalendran et al. (2006), and Goldstein et al. (2010) investigate IPO allocations in return for stock-trading commissions. Reuter (2006) investigate if IPO allocations are tied to stock-trading commissions by studying 1,868 IPOs on NYSE, AMEX and NASDAQ in the period 1996 to Reuter (2006) find a positive relation between stock-trading commissions paid by mutual funds to IPO lead underwriters and mutual fund holdings of IPO shares after the listings. It is concluded that commission payments is a likely reason behind IPO allocations for mutual funds. Reuter (2006) establish a link between IPO allocations and stock-trading commission for mutual funds, but it is not investigated if commission is important for IPO allocations for other investor groups. Nimalendran et al. (2006) study investor trades in the 50 most liquid stocks in the U.S. during the days surrounding IPO allocations. They find that trading volume is positively related to money left on the table in IPOs. It is suggested that this increased trading is done purely to increase stock-trading commissions as payment for IPO allocations. Goldstein et al. (2010) investigate trading by 840 institutional investors around IPOs in the period 1999 to They find that investors both churn stocks and increase the average commission per trade they pay to lead underwriters of upcoming profitable IPOs. All of these papers support the rent-seeking view of IPO allocations. 7

8 3. Testable implications Investors are placed on A, B and C lists by the allocating investment bank before any IPO. 4 Investors from the A list, that applies for shares, are more likely to receive more IPO allocations than investors on the B lists etc. We do not know how investors are placed on the A, B and C lists, but we expect that the lists are related to the pitchbook, the information and the rent-seeking view of IPO allocations. After the book-building/pricing of the issuer, the lead investment bank prepares a book with proposed IPO allocations. This book is given to the board of the issuing company for approval. Anecdotal evidence suggests that most boards approve the proposed allocations without adjustments. Being on the A list of the lead investment bank is therefore very helpful when applying for IPO shares. In this paper we investigate if providing issuer valuation information (the information view), price stability in the period after the listing (the pitchbook view), or stock-trading commission (the rent-seeking view) will place investors on the A, B and C lists of the investment banks before the allocation The rent-seeking view of IPO allocations Rent-seeking is an area that have received allot of attention in the media, but there is limited empirical research on the rent-seeking view. A possible reason for this is the difficulty in obtaining data to investigate rent-seeking. Testing for money transfers from one bank account to another obviously requires very detailed data. The cover-up activities needed to hide transfers as 4 The information about IPO allocation practices is obtained from meetings with former Norwegian investment bankers. 5 The three allocation views are not mutually exclusive within or between IPOs. It is therefore possible that different IPOs are allocated based on different views. It is also possible that different investors are allocated shares based on different views within one IPO. 8

9 legitimate activities can potentially be very creative. In this paper we investigate if IPO allocations are dependent on stock-trading commission as suggested in Reuter (2006), Nimalendran et al. (2006), and Goldstein et al. (2010). We investigate the rent-seeking view by regressing actual IPO allocations on investor generated stock-trading commission before the IPO (while controlling for the other allocation views). If investors are allocated IPO shares because of their stock-trading commission payments, IPO allocations will be positively related to investor commission levels. 3.2 The pitchbook view of IPO allocations The pitchbook view of IPO allocations comes from the sales pitch slides used by investment banks when they compete for IPO contracts; see Ritter (2003). In these sales pitch slides it is usually argued that banks will use their power to divert allocations away from flipping investors and rather allocate shares to more long term buy-and-hold investors. It is further argued that long term buy-and-hold investors will create price stability that will be good for the issuing company in the period after the listing. We investigate the pitchbook view by dividing allocated IPO investors into three groups depending on how they have traded their past IPO allocations. Group one investors flip their shares (sell all shares within the first month after a listing), group two investors hold their shares in the long term (hold some shares longer than six months after a listing), and group three investors are all the remaining investors. 6 We then investigate the pitchbook view by regressing IPO allocations on the number of times investors have been, out of 6 We also show the same results when buy-and-hold investors are defined as investors who keep more than 50% of their allocation for six months. 9

10 all past IPO participations, placed in group one (flipping) and group two (buy-and-hold) by each lead investment (while controlling for the other allocation views). Past buy-hold and Past flipping are calculated for past allocations by each lead investment bank. If investors are allocated IPO shares because of the pitchbook view, flipping will be negatively related to IPO allocations and buy-and-hold will be positively related to IPO allocations. 3.3 The information view of IPO allocations In Benveniste and Spindt (1989) it is argued that investment banks will meet with informed investors to set the IPO offer price before the allocation. Informed investors are rewarded for their pricing service with IPO allocations that are underpriced on average. We expect that pricing information investors participate in IPOs of the same investment banks on a regular basis. We also expect that pricing information investors are likely to be financial institutions with bigger market portfolios. We investigate the information view by regressing IPO allocations on past investor IPO allocations overall, past IPO allocations from the same investment banks, the market value of investor portfolios, and a financial institution dummy (while controlling for the other allocation views). 4 Data description There were 387 companies that list on the Oslo Stock Exchange (OSE) from January 1993 to September New listings are identified from the annual statistics published by the 10

11 OSE. In total, 272 of the 387 companies issued equity to new shareholders as part of the listing. The remaining 115 companies listed as privatizations made by the Norwegian government, cross listings, spin-offs to existing shareholders, mergers, takeovers or directly, without any offerings. The equity offerings to new shareholders are of two types. The first type is described in the listing prospectuses as directed towards international and domestic institutional investors (directed issue). In the directed issues bigger investors are invited to apply for shares. The second type is described in the listing prospectuses as open for all investors to apply for shares (open issue). The 272 companies issue capital in 112 directed and 204 open issues in the sample period. In total, 44 of the 272 companies first issued shares in a direct issue and then later in an open issue. We define both direct and open issues as IPOs. We find the same results when the directed and the open issues are investigated separately. We also find the same results if the issues are pooled together as one issue per company. The OSE requires all shareholders be registered in the Norwegian Central Depository (the VPS) before the listing. The VPS database is comprehensive because it is not possible to list on the OSE without first listing in the VPS. This database includes month-end ownership by all shareholders in all companies that are publicly listed or are considering a listing in the future. The VPS database is unique because it allows for calculation of IPO allocations. IPO allocation data are, in general, very difficult to obtain. We define IPO allocations from the VPS database as the difference in company ownership from the end of the month before to the month of the allocation dates (when shares are transferred). The IPO listing prospectuses made in connection with the listings includes all future allocation/share transfer dates. We also observe that shares are transferred in the specified months in the VPS data. We only investigate IPO allocations to new shareholders. Allocations to existing shareholders (if any) are dropped. 11

12 4.1 IPO allocations To determine IPO allocations, there are three dates that are important in the listing process: when the companies list their ownership records in the VPS, when the companies transfers the IPO shares, and when official secondary trading starts (the listing date on the OSE). These three dates influence at what level of detail we can observe the IPO allocations. When the three dates are in separate calendar months, we are able to calculate comprehensive IPO allocations. We define group one companies as those where the three events are in separate calendar months. We are able to observe comprehensive allocations for 30 directed and 19 open issues. There are 27,880 IPO-investor pairs in these 49 issues. These issues are made by 49 different companies. In total, 26 of these 49 group one companies transfer the IPO shares in the month before the listing month. The remaining 23 companies transfer the shares two or three months before the listing month. We observe company ownership after the IPO shares are transferred, but before trading at the OSE has started. Shares sold over the counter (OTC trading) in the period between the allocation day and the end of the allocation month cannot be detected in the data (for the 26 companies who transfer shares in the month before the listing month). Therefore, investors who buy shares in the OTC market between the allocation day and the end of the allocation month will be treated as allocated investors for these 26 IPOs. However, OTC trading is expected to be a very small phenomenon. It is unlikely that many of the investors to whom IPO shares are allocated will sell these shares in the short period before the listing and potentially lose out on the expected underpricing. The VPS data also includes the ownership before and after 156 other issues (159,712 IPO-investor pairs). These 154 open and the two directed issues are made by 154 different 12

13 companies. 7 In total, 14 of these issues are companies where we observe the directed issue as part of the comprehensive sample. The ownership after these IPOs could be contaminated by either some existing owners or some after-listing trading. Fig. 1 shows when companies list in the VPS, when companies transfer shares, and when companies list on the OSE. The final sample is 189 companies with 205 issues (187,592 investor-issue pairs). 4.2 Variable explanations Table 2 describes the main characteristics for the companies listing on the OSE in the period 1993 to Table A1 provides a definition of all the variables used in the paper. Table 3 describes investor trading characteristics for the IPO investors on the OSE in the period 1993 to The dependent variable IPO allocation is calculated as the number of allocated shares to each investor divided by the total number of shares issued in the IPO (for the 49 companies in group one). Ownership is the same as IPO allocation for group one IPOs, the ownership (minus pre-ipo owners) at the last day of the listing month for group two IPOs, and IPO allocations plus pre-ipo owners for group three IPOs. Commission is the accumulated stock-trading commission generated by the IPO allocated investor in the two years before the IPO allocation date. Commission is calculated as the monthly portfolio turnover times market share prices and a fixed percentage commission rate of 0.075% (the average rate used by 15 of the biggest brokerage 7 For 31 open issues and 80 directed issues we cannot observe the allocation data because the companies list in the VPS database after the shares are allocated (or list on the OSE close to the end of the sample period so that we cannot calculate buy-and-hold). 13

14 houses in Norway). Commission is calculated as buy generated commissions only. 8 Commission below the minimum rate is replaced by the fixed minimum fee for one transaction ($15). Portfolio is the market portfolio value for each allocated investor at xx in the year before the IPO allocation date. Past pair is the accumulated previous number of times (out of possible) the allocated investor has received allocations in a previous IPO by the same lead investment bank. Past buyhold is the accumulated previous number of times the allocated investor has been a buy-and-hold investor as a fraction of all previous IPO participations with the same lead investment bank. This is the number of times the investor has held some IPO allocated shares for more than six months in previous IPOs. All investor trading variables are calculated using all 205 IPOs when allocations in the 49 comprehensive sample IPOs are studied separately. All investor trading variables are also recalculated on a bank by bank basis when the most active bank is studied separately. 5. Empirical results In the empirical analysis we investigate if IPO allocations are dependent on stock-trading commissions (the rent-seeking view), buy-and-hold of past IPO allocations (the pitchbook view), or issuer pricing information (the information view), as explained by Ritter (2003) and Jenkinson and Jones (2004). We do this by regressing investor IPO allocations on past stock-trading commissions, investors past IPO share holding periods, and past IPO allocations between 8 Only buy generated commission is included in Commission to avoid issues related to sell generated commissions from rebalancing portfolios before IPOs. The results are the same when Commission is measured as buy, sell or both. 14

15 investors and investment banks. The regression equation in a standard OLS model is given as (1). Ln (IPO allocation) = a + a Commission + a Portfolio + a Past IPOs + a Past pair ij 0 1 ij 2 ij 3 ij 4 ij + a Past buy-hold + a Past flipping + a Held Cold IPO 5 ij 6 ij 7 ij + a Financial + a Ln Market value + a BV/MV + a Offer price 8 ij 9 j 10 j 11 j 49 T=1 + a VC + a Tech + c IPO dummy + d Year dummy+ e 1 12 j 13 j j t i j=1 t=1 All variables are defined in Table A1. We use two-way clustering to account for error dependencies across firms and years; see Cameron, Gelbach, and Miller (2011). The findings are the same with one-way clustering or standard robust errors. IPO allocation is winsorized at 1% to account for outliers. In robustness we show that winsorizing has no effect on results. Figure 2 show the relation between IPO allocation and the residuals from regression equation (1). There is a positive relation between IPO allocation and the residuals. 5.1 The 49 comprehensive sample IPOs In Table 4 we report the results when Ln(IPO allocation) is regressed on the variables in equation (1) in a standard OLS model for the 49 comprehensive sample IPOs (group one IPOs). The main finding is that IPO allocation is highly dependent on Commission. The point estimate of Ln(IPO allocation) on Commission is 0.7. Investors who increase Commission with $1 million before the allocation will increase IPO allocation by 70%. To control for the pitchbook view of IPO allocations, as described by Ritter (2003) and Jenkinson and Jones (2004), we include Past buy-hold and Past flipping in the regression. Past buy-hold is not related to Ln(IPO allocation). Past flipping is positively related to Ln(IPO allocation). This could be explained by that 15

16 investors with a positive level of Past flipping also have a significantly higher Commission than other investors. It is possible that high Commission investors are allowed to flip to take an immediate advantage of the underpricing. To control for the information view, we include Portfolio, Past IPOs, Past pair, and Financial. We expect that investors who provide pricing information participate on a regular basis with the same underwriters. We also expect that information investor will be regular IPO investors, have big portfolios, and a higher probability of being financial institutions. Ln(IPO allocation) is positively related to Portfolio, Past IPOs, Past pair, and Financial. We also control for company fixed effects, year fixed effects, and company specific variables (Market value, BV/MV, Offer price, VC, and Tech). We find very similar results when retail investors are investigated separately in column 2 of Table 4. Although now the coefficients of Commission on Ln(IPO allocation) is much stronger. The point estimate of Commission on Ln(IPO allocation) is 20. Retail investors who increase Commission with $10,000 will increase IPO allocation by 200%. Ln(IPO allocation) is positively related to Past flipping and unrelated to Past buy-hold. Retail investors are not rewarded for Past buy-hold. Ln(IPO allocation) is positively related to Portfolio, Past IPOs, and Past pair. Pricing information seems to be important for IPO allocation to retail investors. In Colum 3 of Table 4 we investigate institutional investors separately. The coefficient for Commission on Ln(IPO allocation) is Institutional investors with high Commission are rewarded with IPO allocation. There is a positive relation between Ln(IPO allocation) and Past buy-hold. Institutional investors are rewarded for Past buy-hold. There is no relation between Ln(IPO allocation) and Past flipping. There are strong relations between Ln(IPO allocation) and Portfolio, Past IPOs, and Financial. Institutional investors are rewarded for information. 16

17 In column 4 and 5 of Table 4 we drop the pitchbook view variables and the information view variables to control that the results are not driven by correlation between the different views, respectively. In column 6 we drop all control variables. The results remain unchanged. The results are not driven by a correlation between the control variables. The overall findings are that Past buy-hold is important for institutional IPO allocation. Past IPOs and Commission is very important for all IPO allocation. We find support for the pitchbook view for institutional investors. We find strong support for the information view and the rent-seeking view for all investor groups. 5.2 Investigating all 205 IPOs with allocation data The VPS database also include 156 IPOs where allocations might be contaminated by after-listing trading or pre-ipo owners (group two and three IPO allocations). In Table 5 we include the IPO allocations in group two and three in the analysis. The number of investigated IPOs is then 205 and the number of investor-ipo pairs is 187,592. The dependent variable is now labelled Ownership (to avoid confusion as we do not observe comprehensive IPO allocation for the 156 other IPOs). 9 The results remain unchanged. Column 1 of Table 5 includes all investors. The point estimates of Commission on Ln(Ownership) is 1.3. Investors who increase Commission by $1 million will increase Ownership by 130%. Ownership is positively related to both Past buy-hold and Past flipping. Ownership is positively related to Portfolio, Past IPOs, Past pair, and Financial. In column 2 of Table 5 we investigate retail investors separately. The 9 Ownership is the same as IPO allocation for group one IPOs (49 IPOs), the ownership (minus pre-ipo owners) at the last day of the listing month for group two IPOs (151 IPOs), and IPO allocations plus pre-ipo owners for group three IPOs (five IPOs). 17

18 coefficient of Commission on Ln(Ownership) is now 31. There is a positive relation between Ln(Ownership) and Past buy-hold and Past flipping. Ln(Ownership) is unrelated to Portfolio, Past IPOs and Past pair. Retail investors are rewarded with IPO allocation for Commission and Past buy-hold. Retail investors are not rewarded for information. In column 3 of Table 5 we investigate institutional investors separately. The point estimate of Commission on Ln(Ownership) is 0.9. There is a positive relation between Ln(Ownership) and Past buy-hold. There is no relation between Ln(Ownership) and Past flipping. There are strong relations between Ln(Ownership) and Portfolio, Past IPOs, Past pair, and Financial. The 205 IPO sample show support for the pitchbook view for both retail and institutional investors. Past buy-and-hold investors hold more shares immediately after IPOs. There is strong support for the information view by institutional investors. Institutions hold more shares immediately after IPOs because they participate in IPOs on a regular basis. There is no support for the information view for retail investors. There is strong support for the rent-seeking view for all investor groups. Investors who generate large stock-trading commissions hold more IPO shares immediately after the offerings. In regression 4 and 5 of Table 5 we drop the pitchbook view and the information view to control that the results are not driven by correlation between the different views, respectively. In regression 6 we drop all control variables. The results remain unchanged. 5.3 IPO allocations in popular IPOs Investment banks can optionally report the IPO oversubscription to the newspapers after the listing. Oversubscription is defined as the number of applied for shares divided by the 18

19 number of issued shares. Oversubscription is reported for only 117 out of the 205 IPOs in the sample. In the main analysis we assume that all IPOs are oversubscribed. (We assume that there is an excessive demand for shares in all IPOs). Most banks report oversubscription for some IPOs, but not for others. However, one bank publish oversubscription levels for 18 out of 19 public issues. In total, 16 of these 18 IPOs are reported as oversubscribed. This bank acts as the lead underwriter for three private issues (where no oversubscription is reported). Our assumption that IPOs are in general oversubscribed seems reasonable. In Table 6 we investigate the 18 allocations from the one bank with reported oversubscription levels separately. We then use the actual oversubscription as an interaction variable with Commission. Oversubscription takes the value of one for issues that are reported as oversubscribed. The unreported issues are dropped from the analysis. From column 1 it can be seen that the coefficients of Commission and Commission*Oversubscribed on Ln(Ownership)are 0.2 and 1.24, respectively. Ln(Ownership) is positively related to Past flipping. Ln(Ownership) is not related to Past buy-hold. Ln(Ownership) is strongly related to Portfolio, and Financial. Ln(Ownership) is not related to Past IPOs or Past pair. The bank that reports oversubscription the most often does not reward Past buy-hold or Past IPOs investors. Investors who generate large stock-trading commissions own more oversubscribed shares immediately following the new listings. In column 2 and 3 we drop the pitchbook view and the information view, respectively. The results remain unchanged. In column 4 to 6 Oversubscription takes the value of one for IPOs that are more than 2.5 times, more than five times, and more than 19 times oversubscribed, respectively. These are the top 50%, the top 25%, and the top 5% of oversubscribed IPOs, respectively. The relation between Ln(IPO allocation) and Commission*Oversubscribed is stronger with the 19

20 oversubscriptions in the IPOs. Commission becomes more important for IPO allocation when the IPOs are more oversubscribed. We also expect that there will be a positive correlation between expected underpricing and actual underpricing. In Table 7 we interact Commission with the dummy variables Hot and Cold. Hot takes the value of one for the top 25% of underpriced issues. Hot issues are more than 25.5% underpriced (in the comprehensive 49 IPO sample). Cold takes the value of one for the lowest 25% underpriced issues. Cold issues are more than -0.1% overpriced (in the comprehensive 49 IPO sample). From columns 1 and 2 in Table 7 it can be seen that there is a significantly higher relation between Ln(IPO allocation) and Commission in Hot issues. There is also a significantly weaker relation between Ln(IPO allocation) and Commission in Cold issues. Investment banks reward high Commission investors with more Hot IPO allocation. Investment banks also direct Cold IPO allocation away from high Commission investors. It is possible that buy-hold investors are rewarded (and flipping investors punished) only in Hot IPOs instead of in all IPOs. We therefore also study the top quartile of underpriced IPOs separately in column 3 of Table 7. Ln(IPO allocation) is positively related to Commission. There is no relation between Ln(IPO allocation) and Past buy-hold. There is a positive relation between Ln(IPO allocation) and Past flipping. There is a positive relation between Ln(IPO allocation) and Portfolio, Past IPOs, Past pair, and Financial. The interpretation is that investors are allocated underpriced IPO shares because of pricing information and generated stock-trading Commission. Investors are not rewarded with underpriced IPO allocation for Past buy-hold. Investors are also allocated less overpriced IPO allocations because of high stocktrading Commission. We find the exact same results when we investigate all the 205 issues together in columns 4 to 6. 20

21 5.4 Matching allocated investors with non-allocated investors The data only provides us with share allocations and not share applications. As a further robustness measure we therefore also match allocated investors one for one with a non-allocated investor (who we do not know if applied for shares or not) to see if the relation between IPO allocation and Commission is still present when these investors are included in the analysis. In total 11,190 (out of 27,880 investors in group one) are first observed in the data with their IPO allocation. These investors have no previous stock-trading commission, past trading or portfolio size. The remaining 16,690 investors are matched one for one with a non-allocated investor with no IPO participations in the last 12 months on investor type, investor country and number of shares in the portfolio. Among these investors the investor with the closest portfolio size is selected as the matching investor. IPO allocation is then regressed on the same variables as before in a Tobit model (with an upper level cut-off of 100% and a lower level cut-off of zero). The 16,690 matching investors take the value of zero for IPO allocation because they are not allocated IPO shares. In column 1 of Table 8 we show that there is a strong positive relation between IPO allocation and Commission when controlling for non-allocated investors. The point estimates for IPO allocation on Commission is There is a weak relation between IPO allocation and Past buy-hold and Past flipping. There are strong relations between IPO allocation and Past IPOs, Past pair, and Financial. Surprisingly, there is now a negative relation between IPO allocation and Portfolio. In column 2 and 3 we drop the pitchbook view variables and the information view variables, respectively. The results remain unchanged. Investors are allocated IPO shares based on their past commission trading when controlling for non-allocated investors. Some investors are also allocated shares because they participate in the IPO market on a frequent basis. There are also indications that some investors 21

22 may be allocated some IPO shares based on past buy-and-hold of IPO shares. There is strong evidence supporting the rent seeking view when controlling for non-allocated investors. There is some evidence supporting the pitchbook view. There is strong evidence supporting the information view when controlling for non-allocated investors. 5.5 Matching IPOs with IPO Spin-offs As a further robustness test we control that the relation between IPO allocation and Commission is a result of the allocation and not an effect of the new listings. We do this by including the 24 companies where shares are mainly spun off to existing shareholders (rather than to new investors) in the analysis. In many of the 24 spin-offs it is documented in the listing prospectuses (or in the newspapers surrounding the listing) that existing shareholders will get a preference if they apply for shares. We use the immediate after-listing ownership in these companies as a comparison to Ownership. Listing prospectuses are distributed quite thinly in non-ipo listings. For many of the 24 spin-offs it has therefore not been possible to obtain the listing prospectuses. We therefore miss some company control variables for spin-offs. We do, however, include company fixed effects so we do not expect any issues related to missing variables bias. Table 9 show the results when Holdings/issued for the 205 IPOs and the 24 spinoffs is regressed on the same variables as before. 10 We multiply Commission with the interaction term IPO. IPO takes the value of one for the 205 IPO companies and zero for the 24 spin-offs Holdings/issued is the same as IPO allocation for group one IPOs, the ownership (minus pre- IPO owners) at the last day of the listing month for group two IPOs, IPO allocations plus pre- IPO owners for group three IPOs, and the ownership at the last day of the listing month for non- IPOs. 22

23 We use a Tobit model with an upper level cut-off of 100% and a lower level cut-off of 0% on IPO allocation. From column 1 of Table 9 it can be seen that the coefficients for Commission and Commission*IPO on Holdings/issued are 0.19 and 0.12, respectively. The relation between Holdings/issued and Commission is significantly stronger in general IPOs compared to IPO spinoffs. The relation between IPO allocation and Commission is driven by the IPO allocation and not by the new listing. In column 2 and 3 we drop the pitchbook view variables and the information view varaibless, respectively. The results remain unchanged. 5.6 Robustness In the main analysis we winsorize IPO allocation at 1% to adjust for outliers. In column 1 of Table 10 we show the results when no adjustments for outliers are made (IPO allocation is not winsorized). The results are slightly stronger with a coefficient of Commission on IPO allocation of 0.8. The results are not affected by winsorizing. In the main analysis we investigate IPO allocation in Ln (natural logarithm) form. In column 2 we investigate IPO allocation in % rather than in Ln form. The results remain unchanged. In the main analysis we require buy-hold investors to keep all their shares for more than six months to be classified as Past buy-hold. In column 3 of Table 10 Past buy-hold includes investors who have held more than 50% of their shares for six months. There is now no relation between Past buy-hold and Ln(IPO allocation). It is possible that there is a bias in the previous trading variables in the early years of the sample (as investors cannot have previous allocations). Regression 4 drops the first two years of the sample to avoid no previous IPOs bias. The results remain unchanged. 23

24 We also calculate the Variance Inflation Factors (VIFs) to test for multicollinearity. No specifications have a VIF greater than 5, indicating that multicollinearity is not an issue. We also split the sample in two based on high and low investor trading commission (not reported). We then find that the total return is significantly higher for high commission investors compared to low commission investors. 6. Conclusion The main finding of this paper is a strong and robust relation between stock-trading commission generated by investors before IPO allocations and the number of shares investors are allocated in subsequent IPOs. This finding controls for the market portfolio value of the allocated investors, past trading behavior (the pitchbook view) and investor/investment bank relationships (the information view). This finding is also robust to companies who do not list through IPOs and investors who are not allocated IPO shares. We also find that the relation between stocktrading commission and IPO allocations is significantly stronger in hot (oversubscribed or underpriced) IPOs. The meaning of these results are that there is a strong indication that all investors are able to "buy" (hot) IPO allocations with stock-trading commissions. The investors that are the most profitable clients, for the investment banks, are rewarded with the most IPO allocations. It can be argued that bigger investors are likely to apply for more IPO shares. The results are, however, robust to controlling for the market portfolio value of the allocated investors. There is some evidence that support the information view of IPO allocations. Investors who regularly participate in IPOs by the same investment banks are allocated more IPO shares. 24

25 There is not a consistent support for the pitchbook view of IPO allocations. There is not a consistent relation between past IPO share holding periods and current IPO allocations. Investors that hold shares in the long run are not allocated more future IPO shares. The conclusion is that IPO shares are allocated based on the rent-seeking view for all investor types. This finding is consistent with Reuter (2006), Nimalendran et al. (2007) and Goldstein et al. (2010) in that IPO shares are allocated in return for stock-trading commissions. There are some limitations to the study. It is not observed that stock-trading commission is paid from the allocated investor to the investment bank. It is only observed that the commission is generated. Commission is also calculated based on monthly data. This is a lower bound on commission. It is also possible that investors increase trading commission by increasing payments per trade instead of trading volume. We observe trading on a monthly level and are therefore only able to estimate trading volume. The possibility that investors receive allocations for increased payments per trade do, however, work against us and not for our results (as increasing trading volume and payments per trade are substitutes and not compliments). The main contribution to the previous literature is that we are able to combine all existing views on IPO allocations in the same analysis for all investor types. We rank the existing views on how IPO shares are allocated from most to least important. This has not been possible to do before. There is strong evidence supporting the rent-seeking view. There is some evidence supporting the information view and some evidence against the pitchbook view when controlling for the rent-seeking view. We also investigate actual IPO allocations for all the allocated investors in many IPOs. This has also, as far as we know, not been done before. The main practical implication of this study is that investors seem to be able to increase IPO allocations by increasing stock-trading commission before IPOs. Investors can also be able to increase IPO 25

26 allocations by directing trades to investment banks that underwrite many IPOs. Another practical implication is that here should be more regulatory investigations of IPO allocation practices. It seems that the exchange of IPO allocations with stock-trading commission is a widespread practice. For future research it would be interesting to study stock-trading commission that is paid directly to the investment bank by all the allocated investors on a daily basis. 26

27 References Aggarwal, Reena, 2003, Allocation of initial public offerings and flipping activity, Journal of Financial Economics 68, Benveniste, Lawrence and Paul Spindt, 1989, How investment bankers determine the offer price and allocation of new issues, Journal of Financial Economics 24, Binay, Murat, Vladimir Gatchev and Christo Pirinsky, 2007, The role of underwriter-investor relationships in the IPO process, Journal of Financial and Quantitative analysis 42, Bubna, Amit and Nagpurnanand Prabhala, 2008, When bookbuilding meets IPOs, AFA 2008 New Orleans Meetings Paper Available at SSRN: Cameron, C., Gelbach, J., Miller, D., Robust inference with multiway clustering. of Business & Economic Statistics 29, Journal Cliff, Michael and David Denis, 2004, Do initial public offering firms purchase analyst coverage with underpricing?, Journal of Finance 59, Cornelli, Francesca and David Goldreich, 2001, Bookbuilding and strategic allocation, of Finance 56, Journal Cornelli, Francesca and David Goldreich, 2003, Bookbuilding: How informative is the order book?, Journal of Finance 58, Fjesme, Sturla, Laddering in initial public offerings, AFA 2012 Chicago Meetings Paper Available at SSRN: Goldstein, Michael, Paul Irvine and Andy Puckett, 2010, Purchasing IPOs with commissions, Journal of Financial and Quantitative Analysis, forthcoming. Griffin, John, Jeffrey Harris and Selim Topaloglu, Why are IPO investors net buyers through lead underwriters? Journal of Financial Economics 85, Jenkinson, Tim and Howard Jones, 2004, Bids and allocations in European IPO Journal of Finance 59, bookbuilding, Liu, Xiaoding and Jay Ritter, 2010, The economic consequences of IPO spinning, Review of Financial Studies 23, Liu, Xiaoding and Jay Ritter, 2011, Local underwriter oligopolies and IPO underpricing, Journal of Financial Economics, forthcoming. 27

28 Ljungqvist, Alexander and William Wilhelm, 2002, IPO allocations: discriminatory or discretionary?, Journal of Financial Economics 65, Nimalendran, M., Jay Ritter, and Donghang Zhang, 2007, Do today's trades affect tomorrows IPO allocation?, Journal of Financial Economics 84, Reuter, Jonathan, 2006, Are IPO allocations for sale? Evidence from mutual funds, Journal of Finance 61, Ritter, Jay, 2003, Differences between European and American IPO markets, European Financial Management 9, Rogers, W., 1993, Regression standard errors in clustered samples, Stata Technical Bulletin 13, Reprinted in Stata Technical Bulletin Reprints 3, Wooldridge, Jeffrey, 2009, Introductory Econometrics A Modern Approach, Canada, South- Western Cengage Learning. 28