On the Negative Relationship between Initial Returns and IPO Volume Romain Bouis* OECD Economics Department

Abstract This paper investigates why initial returns of IPOs (initial public offerings) depend negatively on the contemporaneous number of firms going public. Using a sample of French offers, I find that this negative relationship does not result from an upward adjustment of the offer price; is more significant for offers completed by poorly experienced banks; and vanishes once controlling for retail investors’ demand, which is found to depend negatively on IPO volume. Taken together, these results contradict an explanation based on information spillovers but are consistent with the presence of a downward sloping demand curve at the market level.

*E-mail address: [email protected] The views expressed in this paper are those of the author and do not necessarily reflect those of the OECD or its member countries.

1

I. Introduction It is well documented that more firms tend to go public following periods of high average initial returns on the IPO market (Lowry and Schwert (2002)) and several hypotheses have been offered to explain this positive lead-lag relationship.1 IPO volume and initial returns are, however, not always positively related. As found by Benveniste, Wilhelm, and Yu (2000), new issuers exhibit on average lower initial returns when more companies are going public at the same time. Surprisingly, the IPO literature has not much analysed why contemporaneous IPO volume reduces individual initial returns and there is to date only one explanation of this effect proposed by Benveniste, Busaba, and Wilhelm (2002). Assuming that initial returns reflect information revelation costs borne by the issuers, Benveniste, Busaba, and Wilhelm (2002) claim that underpricing decreases when several firms whose valuations depend on a common (industry) factor are going public simultaneously. At least two reasons, however, lead to reconsider this information revelation costs argument. First, a number of studies minimize the role of information asymmetries in explaining IPO returns, particularly during hot issue periods, showing instead that the demand from sentiment or retail investors is an important driver of initial returns (Derrien (2005), Cornelli, Goldreich, and Ljungqvist (2006), Dorn (2009)). Second, anecdotal evidence suggests that issuers tend to perceive contemporaneous IPO candidates as rivals in attracting investors rather than as a positive external source of learning. 1

Lowry and Schwert (2002) claim for instance that high initial returns are related to positive information learned by

firms during their registration period, explaining why more companies file to go public following periods of high average initial returns. Ljungqvist, Nanda, and Singh (2006) argue that initial returns reflect the optimism of sentiment investors and that more firms are going public when investor sentiment is high.

2

During the IPO process of Bank of China in 2006, the press was instance reported that “the offer [would] have to compete for investors with a couple of already-listed Chinese banks as well as with several planned listings later [in the] year” (“Bank of China’s Planned IPO Will Widen Choice for Investors”, The Wall Street Journal, March 26, 2006) while Fishe and Boehmer (2000) report similar issues of competition for investors between the IPOs of Andover.net and of VA Linux in 2000. In this paper, I consider a new interpretation of the negative relationship between contemporaneous IPO volume and initial returns based on the hypothesis of a downward sloping demand curve at the market level.2 It is argued that with a fixed demand of investors in the short run, a rise in the number of contemporaneous IPOs leads, on average, to a decline in initial returns because of a lower post-issue market price through a lower demand of investors per issue, rather than because of an upward adjustment of the offer price, as predicted by the information revelation cost hypothesis of Benveniste, Busaba, and Wilhelm (2002). I conduct a series of tests to discriminate between the two competing hypotheses by using a sample of 433 IPOs completed on the French stock market between 1996 and 2006. This latter market offers an ideal testing ground as the IPO process in France is relatively short and, unlike in the U.S., most issuers never revise their price range, allowing to better control for any information spillovers that may influence the pricing of the IPO. Importantly, the presence of a selling tranche exclusively opened to retail investors allows collecting precise information on the demand of investors. Results of the tests tend to invalidate the information revelation costs argument while 2

Following Braun and Larrain (2009), and in contrast with Shleifer (1986), this downward sloping demand curve is

here defined at the market level in the sense that a supply shock from a group of firms may affect the stock price of other companies of the market.

3

lending support to the downward sloping demand curve hypothesis. First, I find that firms do not increase their offer price during the pre-IPO process when the number of offers completed in the weeks before the issue is larger. Contemporaneous IPO volume has even a significant negative impact on the offer price adjustment in some specifications. Second, the negative effect of contemporaneous IPO volume on initial returns does not result from an upward adjustment of the price range as the number of IPOs completed after this price range is set is a significant determinant of IPO returns, even when controlling for the number of offers completed before the price range is set. Third, the negative relationship between initial returns and contemporaneous IPO volume is more significant for firms taken public by low experience investment banks than for other firms. This result contradicts the predictions of Benveniste, Busaba, and Wilhelm (2002) that given their market power, top investment banks are in better position to impose a sharing out of information revelation costs among contemporaneous IPOs. It is, conversely, consistent with the argument that IPOs experiencing a decline in investors’ demand as competition gets tighter are essentially those completed by poorly experienced banks as these latter are less successful than experienced underwriters in marketing the offers. Finally, I analyse the relationships between the number of contemporaneous IPOs, the initial returns and the oversubscription ratios of retail investors. Estimates show that contemporaneous IPO volume has a negative impact on the demand from retail investors observed for each issue while retail investors’ demand affects initial returns positively. Importantly, when the demand from retail investors is included in the regressions of initial returns, these latter do not depend anymore on contemporaneous IPO volume. Results therefore suggest that contemporaneous IPO volume affects individual initial returns through its effect on retail investors’ demand and not because of a sharing of information revelation costs among issuers.

4

Benveniste, Ljungqvist, Wilhelm, and Yu (2003) have already analysed empirically the negative effect of IPO volume on individual initial returns. My approach, however, differs from theirs as I consider contemporaneous offers observed in the few weeks around the IPO date (as Benveniste, Wilhelm, and Yu (2000)) rather than offers from the same industry as the IPO of interest which filed several months before.3 For this reason, although my findings contradict the argument that the negative relationship is due to a sharing of information production costs among contemporaneous issuers, they do not challenge the view that underwriters gather information from firms already traded on the market as evidenced by Benveniste, Ljungqvist, Wilhelm, and Yu (2003). Banks of my sample tend indeed to use public valuations of firms from the same industry as the IPO of interest when pricing the issue by applying, for instance, the so-called multiples valuation method. These comparable firms, whose names are listed in the IPO prospectuses, have, however, generally gone public for several months, or even for several years at the time the issue is valued by the bank and must therefore be distinguished from firms going public in the few weeks surrounding the offer considered here and which do not necessarily operate in the same industry.

The rest of the paper proceeds as follows. Section II presents the two competing hypotheses to explain the negative relationship between initial returns and IPO volume and testable predictions to discriminate between these two hypotheses. Section III describes the institutional features of the French IPO market. Section IV details the data and the variables used 3

More precisely, Benveniste, Ljungqvist, Wilhelm, and Yu (2003) consider in their regressions of initial returns the

number of IPO fillings from the same industry as the offer of interest observed during the first 180 days of an industry IPO wave. The authors find this variable to be even more significant when lengthening the window to 360 days.

5

in the empirical analysis. Section V presents and discusses the econometric results. Section VI concludes.

II. Hypotheses and Testable Predictions This section introduces the two competing hypotheses to explain the negative relationship between initial returns and IPO volume: the information revelation cost hypothesis and the downward sloping demand curve hypothesis. It then presents a set of testable predictions to discriminate between these two hypotheses. A. The Information Revelation Cost and the Downward Sloping Demand Curve Hypotheses The initial return is traditionally measured as the performance of the IPO between the offer price and the close of the first trading day. Its decline can therefore result from an increase in the offer price and/or from a lower post-issue market price. The information revelation cost hypothesis predicts that the negative relationship between initial returns and IPO volume results from an increase in the offer price. This hypothesis is based on the Benveniste and Spindt’s (1989) seminal paper according to which the investment bank underprices the issue to encourage informed investors to truthfully reveal their private valuations about the IPO. Because the valuation of a firm depends on a specific (idiosyncratic) factor but also on a factor common to firms of the same industry (industry factor), information revelation costs decrease with the number of IPOs that belong to the same industry (Benveniste, Busaba, and Wilhelm (2002)). The information revelation cost hypothesis therefore predicts that the issuer and its underwriter can price the offer with a lower discount when more firms are 6

going public, translating into a lower initial return. An alternative explanation of the negative relationship between initial returns and contemporaneous IPO volume is based on the hypothesis of a downward sloping demand curve at the market level coupled with competition between investment banks to attract investors. In the presence of divergence of opinion among potential investors and of short sales constrains, the demand curve for stocks slopes down and the market price depends on the number of optimistic investors willing to acquire the stock (Miller (1977)).4 In general, nothing ensures that potential investors have at any time perfect information on the existence of all the stocks offered on the market, especially when these investors are infrequent traders like retail investors and when the stocks are issued in an IPO, i.e. by a company relatively little known to the public. As argued by Merton (1987), under these conditions the participation of retail investors depends on the efforts of promotion of investment banks as the marketing efforts increase the number of potential investors by ensuring the diffusion of information on the existence of the stock. Cook, Kieschnick, and Van Ness (2006) notice that by doing this, investment banks also increase the divergence of opinion as their efforts of promotion mainly attract retail investors, that is investors whose valuations are more likely to be higher than the average market valuation, at

4

Recent studies point to the presence of significant short sales in the first days following the IPO, contradicting the

hypothesis of constraints on short selling (see e.g. Edwards and Hanley (2010)). Even if short sales are feasible for new issues, noise trader risk (De Long, Shleifer, Summers, and Waldmann (1990)) and other limits to arbitrage may, however, deter investors from engaging in massive short selling that would be necessary to drive stock prices to their fundamental value.

7

least during hot issue periods.5 Over a given time interval, sentiment investors are in limited number and banks probably compete in their activity of marketing towards these investors to maximize the proceeds of the offers. If the offer price partially reflects the demand from sentiment investors, the initial return depends positively on this demand. The negative relationship between initial returns and contemporaneous IPO volume can then be interpreted as the result of a lower secondary market price due to tighter competition between banks to promote the offers towards sentiment investors. B. Testable Predictions According to the information revelation cost hypothesis, the investment banks and the issuers can adjust the offer price upwards by observing the pricing of recent offers, leading to a decrease in initial returns. The number of contemporaneous IPOs should therefore have a positive impact on the offer price adjustment. Conversely, the downward sloping demand curve hypothesis does not necessarily imply an effect of contemporaneous offers on the offer price, as the decrease in initial returns is expected to work through a lower post-issue market price. It is, however, incompatible with an upward adjustment of the offer price following higher IPO volume and may even predict an opposite effect: as more firms compete for investors’ attention, issuers should face on average a lower demand from sentiment investors and revise their offer price downwards. Because the adjustment is only partial (Hanley (1993)), initial returns still depend on investors’ demand. This leads to the following prediction for the downward sloping demand curve hypothesis to hold: 5

In line with this conjecture, Cook, Kieschnick, and Van Ness (2006) find that the participation of retail investors to

the IPO, the adjustment of the offer price, and the initial return, depend positively on the marketing efforts of the banks.

8

Prediction 1 Contemporaneous IPO volume and the offer price adjustment should not be positively related. IPO candidates may also adjust the offer price range (and not only the definitive offer price) in response to information revealed in recent offers. If the negative effect of IPO volume on initial returns is the result of such an adjustment, the return of the offer measured from the midpoint of the offer price range to the post-issue market price should depend on the number of IPOs preceding the setting of the range, but should not depend on the number of IPOs completed after this price range is set. The two IPO volumes (the one observed before and the one observed after the setting of the price range) being correlated, each one is likely to explain negatively the return of the offer. If the decrease in the return is due to a higher price range, the number of IPOs completed after the determination of the price range should, however, be insignificant in explaining this return once considering the IPO volume preceding the setting of the offer price range. This leads to the second prediction. Prediction 2 The number of IPOs completed after the offer price range is set should explain significantly the return measured from the midpoint of the price range to the market price, even after controlling for the number of IPOs preceding the setting of this price range. According to the results of Cook, Kieschnick, and Van Ness (2006), the initial return is positively correlated to the marketing efforts of the offer made by the investment banks while Gao and Ritter (2010) argue in the context of seasoned equity offerings that “marketing can change the short-run elasticity [of demand] and achieve a higher offer price and post-issue market price”. In addition, Cook, Kieschnick, and Van Ness (2006) find that bank reputation and advertising about the IPOs are positively correlated as top intermediaries are probably more

9

successful in promoting the issues. As competition to promote the offers gets tighter (i.e. as the number of contemporaneous IPOs increases), low experience banks may face more difficulties in attracting investors than do high experience intermediaries. The volume of contemporaneous issues should therefore affect more negatively the initial returns of IPOs completed by poorly experienced underwriters than the returns of firms taken public by top-tier banks. The information revelation hypothesis predicts exactly the opposite as with “sufficient market power, an investment bank can spread the costs of information production over many firms” (Benveniste, Busaba, and Wilhelm (2002)).6 The third testable prediction therefore concerns the impact of contemporaneous IPO volume on initial returns depending on the experience of the investment bank. Prediction 3 The negative relationship between initial returns and IPO volume should be more significant for IPOs completed by low experience investment banks. In line with the finding of Barber and Odean (2008) that retail investors are more likely to buy “attention-grabbing stocks” than are institutional investors, Cook, Kieschnick, and Van Ness (2006) find that the efforts of promotion of investment banks explain the buying volume of retail investors in the first days of trading. If investment banks compete to attract retail investors, on average, the demand of retail investors observed for each offer should decrease with the number of contemporaneous IPOs. Note, however, that the presence of a negative relationship between retail investors’ demand and contemporaneous IPO volume is not a sufficient condition to discriminate between the information revelation cost hypothesis and the downward sloping 6

Although contemporaneous IPOs considered here also include companies taken public by other banks, the

information revelation costs hypothesis implies that initial returns of firms taken public by experienced underwriters are, on average, more sensitive to contemporaneous IPO volume than are initial returns of other IPOs.

10

demand curve hypothesis. In periods of high contemporaneous IPO volume, investors may be less keen on participating to an IPO because this issue is on average less underpriced (due to information spillovers from contemporaneous offers). The information revelation cost hypothesis predicts, however, that initial returns should still depend negatively on contemporaneous IPO volume when considering investor’s demand as an explanatory variable. Under this hypothesis, IPO volume per se – and not retail investors’ demand – affects initial returns through information spillovers. Conversely, under the downward sloping demand curve hypothesis, initial returns should not depend on contemporaneous IPO volume anymore once the demand from retail investors is included in the regressions. Overall, the negative relationship between initial returns and contemporaneous IPO volume can only be understood through a decrease in the demand from retail investors, as stated in Prediction 4 below. Prediction 4 The number of contemporaneous IPOs should impact negatively initial returns through a decrease in the demand from retail investors. III. Institutional Features of the French IPO Market This section provides a brief description of the listing markets and of the selling procedures available to IPO candidates in France. Until February 2005, the French stock market was made of three regulated markets: the Premier Marché, the official blue chip market; the Second Marché, a market dedicated to midcap family companies; and the Nouveau Marché, set up in 1996 to welcome small growth companies in high-tech sectors. Since then, the French stock market has experienced a substantial overhaul leading to only one regulated market, called Eurolist by Euronext that encompasses the three above-mentioned regulated markets. Companies quoted on this single 11

regulated market are classified into three compartments depending on their capitalization: compartment A includes large caps (more than 1 billion euros), compartment B comprises mid caps (between 150 million and 1 billion euros), and compartment C small caps (market capitalization of less than 150 million euros). Finally, in April 2005, Euronext launched Alternext, a market with listing requirements and operating methods specially designed for small and midsized companies. As of December 2011, 654 companies were listed on Eurolist Paris (one fifth on compartment A, one fourth on compartment B and more than half on compartment C) and 175 companies on Alternext. Firms wishing to list their shares on one of these markets can choose between several selling methods: the offre à prix minimal (the minimum-price offer), the offre à prix ferme (the fixed-price offer), the offre à prix ouvert (the open-price offer), the placement garanti (guaranteed full underwriting) and the cotation directe (admission to trading). The offre à prix minimal (minimum-price offer) resembles a classical auction procedure. The investment bank and the issuer set a minimum offer price that is announced to investors in the few days preceding the IPO. Retail and institutional investors submit orders by specifying the number of shares and the price of acquisition. The day before the IPO, the bank centralises all orders, enabling it to observe the demand curve of the stock, on the basis of which it decides of the level of the offer price. Only orders with a price comprised between the offer price and a maximum price set by the bank and Euronext, are served on a pro rata basis. The minimum-price offer procedure was largely used in the nineties but has now virtually disappeared. Since 2000, only 3 offers used this auction procedure versus more than 60 between 1996 and 1999. In the offre à prix ferme (fixed-price offer), the bank and the issuer decide of an offer price and investors submit bid orders at this fixed price. An indicative price range can also be announced in the preliminary prospectus. Also in some cases, two blocks of shares are available 12

to investors, depending on the size of the orders. One block is reserved to small size orders (maximum of 50 or 100 shares) or explicitly to retail investors (block A); another block is opened to institutional investors and to orders of any size (block B). Orders submitted in block A enjoy a preferential service rate compared to orders of the second block. The placement garanti (guaranteed full underwriting) corresponds to the U.S. bookbuilding procedure. Once the investment bank has determined an initial price range with the issuers, it promotes the offer towards institutional investors through the road show and solicits bidding intentions (price and quantity) of these investors. The offer price is set on the basis of information communicated by investors and shares are allocated in a discretionary manner. The cotation directe (admission to trading), which has been used increasingly in recent years on Alternext is an auction procedure exclusively reserved to institutional investors (private placement). It is hence similar to the U.S. bookbuilding procedure. Finally, the offre à prix ouvert (open-price offer) is the most recent selling method. Introduced in 1999, this procedure aims at better taking into account the demand of retail investors in the setting of the offer price. A few days before the IPO, the bank and the issuer determine a price range in which investors are invited to submit their orders. The day of the IPO, the offer price is set to the lowest limit of the demand served. As in the case of some fixed-price offers, investors can submit orders in two different blocks of shares, block A and block B, depending on the size of their orders and/or their type (institutional or individual). The French IPO market is an ideal testing ground for the empirical predictions presented in Section II. First, the length of the IPO process is much shorter in France than in the U.S. The price range and the size of the issue are decided on average about two weeks before the IPO date 13

while these latter are announced on average three months before the issue date in the case of U.S offers. This allows a better control for any information spillovers that may arise from contemporaneous offers than with a sample of U.S. offers for which the price range is sometimes revised several times during the registration period. Second, most French issues combine the bookbuilding procedure, which is exclusively reserved to institutional investors, with a so-called market procedure (the fixed-price offer or the open-price offer) in which a fraction of the shares (A orders) is explicitly reserved to retail investors. This feature enables to collect the exact demand of shares from retail investors and to test more directly for the hypothesis of a downward sloping demand curve. IV. Data and Methodology A. The Sample Selection The sample is composed of firms going public on the French stock market between February 1996 and December 2006. I only consider IPOs completed on small- and mid-cap markets (that is the on Second Marché and the Nouveau Marché until 2005 and on Alternext and Eurolist compartments B and C after this date). IPOs completed on the Premier Marché and after 2005, on the Eurolist compartment A are less numerous and much bigger than other offers. Some of these offers also result from nationalisations. I also exclude market transfers and dual listings (in France and abroad). This leaves me with a final sample of 433 offers. For each IPO, information on the characteristics of the offers comes from the preliminary and the definitive prospectuses. I collect the IPO date, the name of the investment bank, the selling procedure, the preliminary and definitive number of shares issued (excluding the overallotment option), the founding date of the company, the number of shares initially reserved to 14

the market procedure (for the open-price offers and the fixed-price offers associated with a placement), the offer price range (for the placements, the open-price offers and the fixed-price offers), the minimum offer price (for the auctioned offer), the proportion of shares sold by the insiders (secondary shares) and finally, the definitive offer price and its setting date. Information related to the results of the IPOs, i.e. the demand of retail investors (block A) and the respective allocations for fixed-price and open-price offers, is collected from Euronext. In case the market disequilibrium at the IPO date is too severe, preventing the shares from listing in a predefined variation threshold (usually 20%), Euronext suspends the transactions and the equilibrium price can be reached several days after the date of the IPO (on average about two days later). For this reason, I collect from Datastream the first equilibrium post-issue closing price, which does not necessarily correspond to the closing price at the end of the first trading day.7 B. Explanatory Variables I consider the following explanatory variables in the econometric analysis. Stock market return The buy-and-hold return of the SBF 250 index measured over the three months preceding the IPO is used as a proxy for market conditions prevailing around the IPO date. As documented in several studies (Lowry and Schwert (2002) or Derrien (2005), among others), this variable is expected to have a positive influence on initial returns. 7

For delisted companies absent from the category “Dead” of Datastream, the first equilibrium price is hand-

collected from financial newspapers.

15

Contemporaneous IPOs The number of contemporaneous IPOs is expected to have a negative effect on initial returns. This variable is defined as the number of IPOs completed across a range of n weeks before to one week after the offering of interest, with n varying from 2 to 8.8 Issue size The size of the issue, which is likely to be positively correlated with the size of the firm, is expected to have a negative impact on initial returns. Smaller firms are likely to be more difficult to market towards investors because of greater information asymmetries and a more risky profile. The issue size is measured as the middle of the initial price range times the expected number of shares offered (in millions of 2000 euros). Firm’s age The age of the company at the issue date is also expected to be negatively related to initial returns as the valuation of older firms is surrounded by less uncertainty, divergence of opinion or asymmetries of information. Secondary shares As documented by Huyghebaert and Van Hulle (2006), small growth firms tend to issue primary shares while more established ones are more likely to issue secondary shares. The ratio of secondary shares to the total number of shares offered (in percent) may then be negatively

8

I focus here on the number of contemporaneous offers. Using instead total proceeds of contemporaneous IPOs

yields qualitatively similar results.

16

related to initial returns as the valuation of small growth firms is more difficult and/or prone to a higher divergence of opinion than the valuation of mature firms. Secondary share sales may also affect initial returns by conveying a negative signal to investors about the information held by existing shareholders. Nascent industries dummy The valuation of firms from nascent industries is also prone to more divergence of opinion and to higher information asymmetries than other firms, resulting on average in higher initial returns. I therefore consider a dummy variable equal to one if the company taken public belongs to a nascent industry, and zero otherwise.9 Offer price adjustment The adjustment of the offer price from the middle of the price range for the fixed-price offers, the open-price offers, and the placements – and from the minimum price for the auctioned offers – should be positively related to initial returns as first shown by Hanley (1993). Auctioned and open-price offers dummies Derrien and Womack (2003) find that IPOs completed with a minimum-price offer procedure exhibit lower initial returns as this selling method allows considering more

9

Nascent industries are those with the following ICB (Industry Classification Benchmark) codes: 2737 (Electronic

equipment), 4535 (Medical equipment), 6535 and 6775 (Telecommunications), 9533, 9535, and 9537 (Software and Computer services), 9572, 9574, 9576, and 9578 (Technology hardware and equipment), 4573 and 4577 (Pharmaceuticals and Biotechnologies). IPOs with an Internet related business (mainly in media) are also considered as nascent offers.

17

information on investors’ demand than fixed-price offers. Likewise, the open-price procedure has been introduced in 1999 with the aim of better incorporating the demand of retail investors in the setting of the offer price. Dummies for auctioned offers and open-price offers are therefore expected to have a negative effect on initial returns. Number of days between the offer date and the date of first equilibrium price Oversubscribed offers are more likely to see their opening price outside a pre-defined price range (usually set to ± 20% of the offer price), leading Euronext to suspend the quotation of the shares. In this latter case, the new reference price becomes the threshold previously reached and trading may start again several days after the IPO date, depending on the severity of the market disequilibrium. I therefore conjecture that the number of days elapsed from the IPO date to the date an equilibrium price is reached for the first time on the secondary market is positively related to initial returns. Postponed offer dummy Some firms are going public after having withdrawn their offer several weeks or months before. These situations appear in general because the demand from investors is too weak in the first IPO attempt. The price range and also sometimes the number of shares offered are then revised downwards. I expect these postponed offers to exhibit lower initial returns, despite a downward revision in their price range. Finally, all regressions include dummy variables for the market of listing (market fixed effects) and fixed year effects.

18

C. Descriptive Statistics Table 1, Panel A reports the number of offers per year as well as the number of IPOs in nascent industries. Panel B shows the evolution in the use of the selling mechanisms. A striking trend is the decline in auctioned IPOs (minimum-price offer) coupled with the rise in the number of placements associated with open-price offers. According to Derrien and Womack (2003), auctioned IPOs have a lower fees rate than other procedures and would not be lucrative enough for banks in a context where the proportion of smaller issues has risen. The declining use of the minimum-price offer may also result from the growing importance of analysts’ recommendations in IPOs (Degeorge, Derrien, and Womack (2007)). Panel C reports descriptive statistics of the variables related to the market conditions in which IPOs take place and descriptive statistics of the initial returns. The average and the median market returns recorded in the three months preceding the offers are positive (+6.3% and +7%, respectively), as firms are in general taken public in periods of rising market valuations. The number of contemporaneous IPOs has a relatively high standard deviation, reflecting the well-known cyclical nature of the IPO market. Finally, the average initial return of the sample (14.7%) is in line with the levels generally reported for advanced economies. [INSERT TABLE 1 ABOUT HERE] V. Empirical Results Before presenting the results of the tests of the predictions set out in Section II, I report in Table 2 the ordinary least squares (OLS) estimates of the relationship between the initial return and the number of contemporaneous IPOs defined over various horizons. As can be seen from this table, the number of contemporaneous offers has a negative impact on initial returns, with a 19

statistically significant coefficient at the 1 or 2% level, depending on the time span used in the definition of contemporaneous offers (columns (1) to (7)).10 The economic impact of IPO volume on individual initial returns is non-negligible. For example, a one-standard deviation increase in the number of offers completed from four weeks before to one week after the IPO of interest (+7.6) leads on average to a decrease in initial returns of about 4.3 percentage points, i.e. to a decrease in initial returns of more than one quarter of its average value. Considering the number of offers completed two weeks after the issue date in the definition of contemporaneous IPO volume does not improve the explanatory power of this variable (columns (8) and (9)). Conversely, contemporaneous IPO volume is always more statistically significant when it includes the number of offers completed one week after the issue date than when it only incorporates the number of IPOs completed in previous weeks (results not reported). [INSERT TABLE 2 ABOUT HERE] The control variables used in Table 2 are statistically significant with the expected signs, except the size of the issue, which has a positive coefficient, and the proportion of secondary shares which is not significant.11 Older companies, firms taken public with an auction mechanism or with an open-price procedure and those that have been previously postponed

10

I also consider in unreported regressions the effects of the number of contemporaneous offers operating in the

same sector (according to ICB codes) as the IPO of interest. The estimated impacts are negative but not statistically significant at conventional levels, suggesting that initial returns react to asset supply shocks observed at a broader level than the IPO sector. 11

This latter result is consistent with Brau, Li, and Shi (2007) who find for a sample of U.S. IPOs that secondary

shares sales are not significantly related to initial returns.

20

exhibit lower initial returns, while the stock market return before the offer, the offer price adjustment, the nascent industry dummy, and the number of days elapsed from the offer date to the first equilibrium market price (a proxy for excess demand) are positively related to the initial returns. A. Is the Negative Relationship Between Initial Returns and Contemporaneous IPO Volume the Result of an Increase in the Offer Price during the Registration Period? This section tests the first empirical prediction that contemporaneous IPO volume and the offer price adjustment are not positively correlated. Table 3, Panel A shows the descriptive statistics of the offer price adjustment for all the issues completed with an open-price offer, with a fixed-price offer, or with a placement alone and that announced an offer price range in their preliminary prospectus. I excluded from this sample the 26 IPOs that have been previously postponed because of a too weak interest from investors in their first attempt to go public and which have adjusted their price range downwards. Most of these postponed offers occurred in June and July 2000, after the burst of the dot-com bubble. Among the 298 remaining issues, only 5 observations have a definitive offer price set outside the indicative price range. As pointed out by Cornelli, Goldreich, and Ljungqvist (2006), unlike in the U.S., the offer price of European offers is almost always set inside the indicative offer price range. This is partly due to the fact that the offer price range is announced later than in the U.S. Because of this feature, I estimate the determinants of the offer price adjustment by using Tobit and Logit models that are left- and right-censored, respectively by the lowest and the highest possible revisions inside the price range, that is

P − PM PL − PM and H , where PM is the middle of the price range, PL the lower PM PM

21

bound and PH the higher bound.12 The explanatory variables are the buy-and-hold return of the SBF 250 index over the three months preceding the setting of the definitive offer price, the size of the offer, the age of the issuer, a dummy variable equal to one if the firm operates in a nascent industry, a dummy variable equal to one for open-price offers, and the number of firms completed in the weeks preceding the setting of the offer price (in logarithms). Columns (1) and (2) of Table 3, Panel B show Probit estimates of the determinants of the probability that the offer price is set at the upper limit of the price range (OP = PH). Columns (3) and (4) and columns (5) and (6) report respectively Tobit and Logit estimates of the determinants of the offer price adjustment from the middle of the price range (i.e.

OP − PM ). Results indicate PM

that both the probability that the issue is priced at the upper limit of the indicative price range and the size of the adjustment depend positively on stock market conditions over the preceding months. IPOs from nascent industries are also more likely to be priced upwards and exhibit a larger positive adjustment of their offer price. Contemporaneous IPO volume (measured over a range of 3 weeks or 6 weeks before the IPO of interest), however, does not lead to higher offer price revision as predicted by the information spillovers hypothesis.13 The coefficient is even negative at the 6% significance level when the regression is estimated with a logistic censored model (column (5)). I also consider the number of IPOs completed up to 9 weeks before the setting of the definitive offer price (results not reported) but the coefficient of this variable is 12

I exclude from the estimation sample the five offers with an offer price set outside the price range.

13

I also analyse the determinants of the probability that the issue is priced at the lower limit of the price range

(results not reported). The contemporaneous IPO volume has a non-significant positive coefficient, while the market return and the dummy variable for nascent industries explain negatively the probability that the offer price is set at the lower limit of the price range.

22

always negative. This suggests that the issuers revise their offer price downwards by a larger extent when contemporaneous IPO volume is higher, consistent with the idea that IPOs may face on average a lower demand for their stocks when more firms are simultaneously going public over a short time interval. In line with Prediction 1, results therefore clearly indicate that the negative relationship between initial returns and contemporaneous IPO volume is not due to an increase in the offer price during the registration period. [INSERT TABLE 3 ABOUT HERE] The negative relationship between initial returns and IPO volume may nevertheless result from a higher indicative price range as issuers get a more accurate prior of their market value in periods of higher contemporaneous IPO volume. In this case, the number of offers preceding the setting of the price range should be more significant in explaining the IPO return than should the number of offers completed after the setting of this price range. The next section tests this prediction.

B. Is the Decrease in the Initial Return as Contemporaneous IPO Volume Increases Due to an Increase in the Offer Price Range? Table 4 shows estimates of the determinants of the return of the IPOs from the middle of the offer price range to the first equilibrium closing price on the secondary market. This return does not depend by definition on the offer price. Yet, as can be seen from columns (1) and (2), the number of contemporaneous offers has a significant negative impact on this return, further supporting the previous finding that the negative relationship between initial return and IPO volume is not the result of an upward adjustment of the offer price during the registration period. As developed in Section II, this negative relationship may, however, result from an increase in 23

the price range. [INSERT TABLE 4 ABOUT HERE] Information on the exact date the price range is set by the issuer and the investment bank is in general not available. It is, however, possible to observe the date this price range is made public for the first time in the preliminary prospectus. For each offer announcing a price range in the preliminary prospectus (i.e. virtually all issues except auctioned offers), I collect the number of offers completed in the few weeks before the price range is announced (i.e. the supposed date of the price range setting), as well as the number of IPOs completed from this date to one week after the offer date of the IPO of interest. Columns (3) to (6) report estimates of the effects of the IPO volumes observed before and after the price range is announced for the first time by the issuer and the underwriter. Both the number of offers observed before the price range is publicly announced and the number of offers completed from this announcement date to one week after the IPO date have a significant negative impact on the return of the offer when considered separately (columns (3) and (4)). If the negative relationship between initial returns and IPO volume is due to an upward adjustment of the price range, the number of IPOs completed after this price range is announced (and until one week after the IPO of interest) should be a poor predictor of IPO returns when controlling for IPO volume observed before the announcement. Results reported in column (5) indicate this is not the case: IPO volume observed after the price range is announced still explains negatively the IPO return when considering IPO volume before the announcement. This result holds true whatever the number of weeks over which IPO volume prior to the price range announcement is calculated.

24

It has been assumed so far that the price range is set the date it is publicly announced in the prospectus. This assumption, however, clearly overstates the true setting date. It appears from a few observations for which information is publicly available, that the issuer and the bank choose on average the price range in the week before announcing it in the prospectus.14 When the price range is assumed to be set around one week before its public announcement date on the prospectus, the number of offers completed before this supposed setting date is no more significant in explaining the IPO return (column (6)), while IPO volume observed after the announcement date of the price range in the prospectus remains significant. Again, this finding is qualitatively the same whatever the number of weeks considered in measuring IPO volume before the announcement of the price range. The results presented in this section therefore support Prediction 2 by suggesting that IPO volume observed after the setting of the price range – rather than IPO volume observed before this setting – plays a significant role in explaining the negative relationship between initial returns and contemporaneous IPO volume.

C. Does Investment Bank Experience Influence the Negative Relationship between Initial Returns and IPO Volume? In this section, I investigate whether the negative impact of contemporaneous IPO volume on initial returns differs depending on the experience of the investment bank. I measure

14

I am able to collect the exact date the offer price range is set by the issuer and the bank for 52 firms that filed in

2005 and 2006. For this subsample of observations, the issuer and the investment bank decide the limits of the offer price range on average 4 calendar days (with a standard deviation of 6 days) before the date the price range is publicly announced for the first time in the prospectus.

25

the experience or reputation of the investment banks by their market share as lead underwriters of the issues. The market share is defined on the basis of the number of firms taken public or alternatively on the basis of total IPO proceeds and is built either over the whole sample period or from the first year of the sample to the date of the IPO of interest, by controlling for all the mergers and acquisitions that took place in the French banking sector over the period. The rankings obtained with these various measures are roughly the same and allow identifying six lead underwriters as particularly active on the French IPO market. Table 5, Panel A reports the number of firms taken public by these six most active investment banks, by distinguishing the IPOs completed with an auction procedure from other offers. The role of the investment bank in auctioned offers is indeed less important than in bookbuilt offers as there is no firm commitment in the former case (the risk of the operation is mainly borne by the issuer) and because the offer price is set jointly with Euronext after the centralisation of the orders. Panel B shows descriptive statistics of the number of IPOs completed by the fifty other banks acting as leaders of the offers (low experience banks). Panel C presents the results of the tests of equality of means and medians of a set of variables for the two subsamples (low experience versus high experience investment banks). The null hypotheses of equality of means (Anova test) and of equality of medians (Wilcoxon/Mann–Whitney test) of the initial returns of the two groups cannot be rejected, indicating that firms taken public by low experience investment banks do not perform differently in the short term from those taken public by highly experienced banks. This result is in line with previous findings from Degeorge and Derrien (2001) that the reputation of the leader does not influence significantly the initial returns in the case of French IPOs. Conversely, high experience investment banks tend to take firms public in 26

periods of better market conditions (as proxied by the three-month market return before the offer) and in periods with higher contemporaneous IPO volume than do low experience investment banks. Finally, both groups of banks are associated with the same proportion of offers from nascent industries and with issuers of the same age. [INSERT TABLE 5 ABOUT HERE] Table 5, Panel D shows econometric estimates of the determinants of initial returns by including in the first two specifications (columns (1) and (2)) the number of contemporaneous IPOs interacted with a dummy variable equal to one if the firm is taken public by a poorly experienced bank, and zero otherwise. As indicated by the results, the negative relationship between initial returns and IPO volume is more pronounced for firms taken public by poorly experienced investment banks, although the negative coefficient of the interaction term is not statistically significant at conventional levels. Columns (3) to (6) present results of the regressions by splitting the sample on the basis of the experience of the investment bank. Estimates show that the negative relationship between initial returns and IPO volume is statistically and economically more significant for the sample of offers completed by poorly experienced investment banks, in line with Prediction 3. The coefficient for the group of low reputation banks is always significant at a statistically level comprised between 1.5 and 5%, depending on the number of weeks preceding the offer date (columns (5) and (6)), while for highly experienced banks, it is only significant at the 10% level (columns (3) and (4)). Moreover, compared to the sample of firms taken public by experienced underwriters (columns (3)), the negative effect of IPO volume is more than three times larger for the sample of firms taken public by poorly experienced banks (columns (5)). A one-standard deviation increase in the number of contemporaneous offers to IPOs taken public by low reputation banks (+6.8 offers) 27

leads to a decrease in the initial returns of these IPOs of about 7.3 percentage points (for an average initial return of 13.6%), while a one-standard deviation increase in the number of contemporaneous offers to IPOs taken public by high reputation banks (+8.1 offers) is associated with a decrease in initial returns of only 2 percentage points (for an average initial return of 15.5%). Results are therefore consistent with the view that poorly experienced banks are facing more difficulties to sell the IPO shares when competition gets tighter, leading to lower initial returns. An alternative interpretation of the results may be that low experience banks have more to learn from the completion of recent IPOs when pricing the issue than do highly experienced banks. By observing the completion of contemporaneous offers, less experienced banks would get a more accurate estimate of the demand from investors and could therefore set a price range and an offer price closer to the secondary market price. This may explain why the negative relationship between initial returns and IPO volume holds true mainly for issues completed with poorly experienced intermediaries. To test for this conjecture, I estimate a regression similar to the one reported in column (5) of Table 4, where the return of the offer from the middle of the price range to the close of the first post-issue equilibrium price is explained by the IPO volume surrounding the setting of the price range. Estimates (not reported) indicate, however, that the number of offers preceding the setting of the price range does not explain significantly the return of the issues completed by low experience banks once the number of IPOs completed after the setting of the price range is taken into account, therefore invalidating the hypothesis of a learning

28

of the state of investors’ demand by low experience banks.15

D. Is the Negative Relationship between Initial Returns and Contemporaneous IPO Volume the Result of a Negative Effect of IPO Volume on the Demand from Retail Investors? This section further investigates the hypothesis of a downward sloping demand curve at the market level by exploiting information on oversubscription ratios of retail investors available for a sub-sample of fixed-price and open-price offers. Table 6, Panel A shows descriptive statistics of the shares reserved to the market procedure (either a fixed-price offer or an open-price offer) and of retail investor’s oversubscription ratios. Information is available for virtually all fixed-price and open-price offers, leaving me with a sample of 286 offers. More than 20% of the shares issued are on average reserved to the market procedure – the remaining shares being sold through a bookbuilding procedure exclusively opened to institutional investors (except for 19 IPOs completed exclusively with a fixed-price offer). Within the market procedure, a tranche is explicitly reserved to retail investors, who submit the so-called A orders. Retail investors’ oversubscription ratios are therefore defined as the ratios of the number of shares demanded by

15

The negative relationship between initial returns and contemporaneous IPO volume may also be an artifact as

lower quality/lower initial return firms managed by less reputable underwriters may rush to go public in hyped markets when contemporaneous IPO volume is higher. This hypothesis is however not supported by the data as firms going public with low reputation banks do not differ significantly in terms of initial returns from those managed by high experienced intermediaries and are in fact going public in times when contemporaneous IPO volume is lower (Table 5, Panel C).

29

A orders to the total number of shares allocated to these orders.16 On average, the retail investors’ oversubscription ratio is equal to 6.75, with a maximum of 100 reached for the IPO of an Internet company (Sys-Com) in March 2000, while about 25% of observations have a subscription ratio strictly equal to one. Finally, the demand from retail investors in the tranche of A orders represents on average around 160% of the total number of shares offered (with a substantial cross-sectional variation), suggesting that for some offers, the demand from retail investors can have a material impact on the secondary market price. [INSERT TABLE 6 ABOUT HERE] Table 6, Panel B reports results of the regressions of retail investors’ oversubscriptions ratios, depending on the experience of the investment bank. As predicted, the demand from retail investors per offer is on average lower when more companies are simultaneously going public, as evidenced by the negative coefficients on contemporaneous IPO volume. Interestingly, this latter effect is statistically significant only for the sample of firms going public with low experience banks (columns (1) and (2)). Results remain robust to excluding the 19 observations with an oversubscription ratio greater than 20 and to varying the number of weeks over which contemporaneous offers are observed. As discussed in Section II, the above finding of a negative effect of IPO volume on retail investors’ demand per offer is, however, not a sufficient condition to reject the information

16

In the case of open-price offers, investors are invited to place orders inside a price-range announced by the issuer

and the bank. For more than 80% of the issues, retail investors place A orders at the upper-limit of the price range, so that the oversubscription ratios are simply defined as the ratios between the number of shares from A orders and the number of shares allocated to these orders (without considering the price at which orders are submitted).

30

revelation cost hypothesis and to validate the downward sloping demand curve hypothesis. The explanatory power of contemporaneous IPO volume in the regression of initial returns should also vanish once the demand from retail investors is included as an explanatory variable. Regressions of Table 6, Panel C test whether this is the case. First, results indicate that the negative relationship between IPO volume and initial returns is also present for the sub-sample of open-price and fixed-price offers, although it is statistically significant only for IPOs completed with low experience banks (columns (1) and (2)). Second, in line with Derrien’s (2005) result obtained for a smaller sample of issues, retail investors’ oversubscriptions ratios do positively explain initial returns, both for firms taken public by low experience bank and for IPOs completed with high experience banks (columns (3) and (8)). Third and most importantly, contemporaneous IPO volume is no more significant in explaining initial returns once the regressions include the demand from retail investors (columns (4) and (5)). This suggests that contemporaneous IPO volume has only an indirect effect on initial returns by reducing the demand of retail investors per issue but does not have any predicting value in itself for initial returns, as would imply the information revelation costs hypothesis. Results therefore corroborate Prediction 4 that the negative impact of contemporaneous IPO volume works through a decrease in the demand from retail investors, further supporting the downward sloping demand curve hypothesis.

VI. Conclusion This paper provides evidence that the negative relationship between initial returns and contemporaneous IPO volume reflects the presence of a downward sloping demand curve for stocks at the market level rather than a sharing of information production costs among issuers. Empirical analysis indicates that the number of contemporaneous issues has a negative rather 31

than a positive impact on the offer price adjustment while IPO volume observed before the setting of the price range is not a better predictor of initial returns than IPO volume observed after this price range is set. The negative relationship between IPO volume and initial returns besides holds true mostly for issues completed by low experience investment banks. This contradicts Benveniste, Busaba, and Wilhelm’s (2002) prediction that prestigious intermediaries are in better position to impose a spreading of information production costs across issuers but is in line with the argument that top-tier banks can escape competition through successful marketing efforts. Finally, the demand of retail investors per issue, rather than contemporaneous IPO volume per se, appears as a significant determinant of individual initial returns. My findings contribute to a recent literature that analyses how shocks to asset supply influence the prices of other assets. Braun and Larrain (2009) document for instance that IPOs in emerging markets lead to a decrease in the prices of existing assets in the market. While this type of effect is, according to the authors, unlikely to be observed in well-integrated markets as in the U.S., a clustering of IPOs over a short period may, however, represent a sufficiently large supply shock and hence have a significant effect on a particular segment of the market, even in welldeveloped financial markets. The negative relationship between individual initial returns and contemporaneous IPO volume analysed here offers a clear illustration of Braun and Larrain’s (2009) conjecture. Hsu, Reed, and Rocholl (2010) also document a negative impact of IPOs on already listed firms of the same industry in the United States but interpret their finding as the result of a fundamental change of listed companies of the sector rather than as the consequence of a financial assets’ supply effect. Although such fundamental effects may be at work in the negative relationship between initial returns and IPO volume, it is worthwhile to note that

32

contrary to Hsu, Reed, and Rocholl (2010), the negative impact of contemporaneous IPO volume on initial returns is here found to hold even among firms from different activities. This analysis finally adds to a new strand of the literature that considers the services of the investment banks in terms of marketing to generate a demand for the stocks (Cook, Kieschnick, and Van Ness (2006), Chahine, Ljungqvist, and Michaely (2007), and Gao and Ritter (2010)) rather than in terms of certification of the issues or of production of information about the issuers. Relaxing the hypothesis of a flat demand curve and introducing competition issues between investment banks to create a demand for the stocks should improve our understanding of the behaviour of companies and financial intermediaries in IPOs and other corporate events.

33

References Barber, B. M., and T. Odean. “All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors.” Review of Financial Studies, 21 (2008), 785-818. Benveniste, L. M.; W. Y. Busaba; and W. J. Wilhelm. “Information Externalities and the Role of Underwriters in Primary Equity Markets.” Journal of Financial Intermediation, 11 (2002), 61-86. Benveniste, L. M.; A. P. Ljungqvist; W. J. Wilhelm; and X. Yu. “Evidence of Information Spillovers in the Production of Investment Banking Services.” Journal of Finance, 58 (2003), 577-608. Benveniste, L. M., and P. A. Spindt. “How Investment Bankers Determine the Offer Price and the Allocation of New Issues.” Journal of Financial Economics, 24 (1989), 343-361. Benveniste, L. M.; W. J. Wilhelm; and X. Yu. “Evidence of Information Spillovers in the Production of Investment Banking Services.” Working paper, Boston College, Chestnut Hill, MA (2000). Brau, J. C.; M. Li; and J. Shi. “Do Secondary Shares in The IPO Process have a Negative Effect on Aftermarket Performance?” Journal of Banking & Finance, 31 (2007), 2612-2631. Braun, M., and B. Larrain. “Do IPOs Affect the Prices of Other Stocks? Evidence from Emerging Markets.” Review of Financial Studies, 22 (2009), 1505-1544. Chahine, S.; A. P. Ljungqvist; and R. Michaely. “Marketing Financial Claims.” Working paper, New York University (2007). Cook, D. O.; R. Kieschnick; and R. A. Van Ness. “On the Marketing of IPOs.” Journal of

Financial Economics, 82 (2006), 35-61.

34

Cornelli, F.; D. Goldreich; and A. P. Ljungqvist. “Investor Sentiment and Pre-Issue Markets.”

Journal of Finance, 61 (2006), 1187-1216. De Long, J. B.; A. Shleifer; L. H. Summers; and R. J.Waldmann. “Noise Trader Risk in Financial Markets.” Journal of Political Economy, 98 (1990), 703-738. Degeorge, F., and F. Derrien. “Les déterminants à long terme des introductions en bourse : le cas français.” Banque et Marchés, 55 (2001), 8-18. Degeorge, F.; F. Derrien; and K. L. Womack. “Analyst Hype in IPOs: Explaining the Popularity of Bookbuilding.” Review of Financial Studies, 20 (2007), 1021-1058. Derrien, F., and K. L. Womack. “Auctions vs. Bookbuilding and the Control of Underpricing in Hot IPO Markets.” Review of Financial Studies, 16 (2003), 31-61. Derrien, F. “IPO Pricing in “Hot” Market Conditions: Who Leaves Money on the Table?”

Journal of Finance, 60 (2005), 487-521. Dorn, D. “Does Sentiment Drive the Retail Demand for IPOs?” Journal of Financial and

Quantitative Analysis, 44 (2009), 85-108. Edwards, A. K., and K. W. Hanley. “Short Selling in Initial Public Offerings.” Journal of

Financial Economics, 98 (2010), 21-39. Fishe, R. P. H., and E. Boehmer. “Do Underwriters Encourage Stock Flipping? A New Explanation for the Underpricing of IPOs.” Working paper, University of Richmond and University of Oregon (2000). Gao, X., and J. Ritter. “The Marketing of Seasoned Equity Offerings.” Journal of Financial

Economics, 97 (2010), 33-52. Hanley, K. W. “The Underpricing of Initial Public Offerings and the Partial Adjustment Phenomenon.” Journal of Financial Economics, 34 (1993), 231-250.

35

Hsu, H.-C.; A. V. Reed; and J. Rocholl. “The New Game in Town: Competitive Effects of IPOs.” Journal of Finance, 65 (2010), 495–528. Huyghebaert, N., and C. Van Hulle. “Structuring the IPO: Empirical Evidence on the Portions of Primary and Secondary Shares.” Journal of Corporate Finance, 12 (2006), 296-320. Ljungqvist, A. P.; V. Nanda; and R. Singh. “Hot Markets, Investor Sentiment, and IPO Pricing.”

Journal of Business 79 (2006), 1667-1702. Lowry, M., and W. B. Schwert. “IPO Market Cycles: Bubbles or Sequential Learning?” Journal

of Finance, 57 (2002), 1171-1200. Merton, R. C. “A Simple Model of Capital Market Equilibrium with Incomplete Information.”

Journal of Finance, 42 (1987), 483-510. Shleifer, A. “Do Demand Curves for Stocks Slope Down?” Journal of Finance, 41 (1986), 579590.

36

Table 1: Descriptive Statistics of IPOs and Market Conditions Table 1 reports the annual number of IPOs per sector (Panel A) and per selling procedure (Panel B) as well as descriptive statistics of market conditions surrounding the offers (SBF 250 buy-and-hold return over the three months preceding the IPO date and the number of IPOs completed from four weeks before to one week after the IPO) and of initial returns (Panel C).

Panel A: Number of IPOs per Year Year of Offer 1996 1997 1998 1999 2000 2001 2002 Total Nascent industries

40 13

54 16

108 39

57 31

60 40

12 4

Panel B: Selling Methods per Year Year of Offer 1996 1997 1998 1999 2000 2001 Selling method Auction procedure Fixed-price offer Bookbuilding and fixedprice offer Bookbuilding and open-price offer Bookbuilding only

2004

2005

2006

6 1

5 1

25 12

66 30

200 2

2004

2005

2006

15 5 12

16 9 23

28 8 62

4 0 21

0 0 19

1 0 1

0 0 0

0 0 0

2 0 1

0 0 5

0

0

0

30

41

10

6

5

20

50

8

6

10

2

0

0

0

0

2

11

Panel C: Market Conditions and Initial Returns Mean Median Std. Dev. Min. Stock market return 6.25% 7.04% 9.30% -30.97% Contemporaneous offers 9.41 7 7.61 0 Initial return 14.70% 6.07% 28.47% -25.49%

37

Max. 30.93% 36 240.91%

Table 2: Determinants of Initial Returns and Contemporaneous IPO Volume Table 2 presents OLS estimates of the effect on initial returns (in %) of the number of contemporaneous offers over different periods (in log). Control variables are the buy-and-hold return of the SBF 250 index over the 3 months preceding the offer, the initial size of the offer (log of constant euros), the age of the firm at the IPO date (log of number of years), the adjustment of the offer price from the middle of the price range to the final selling price, a dummy variable equal to one if the firm is from a nascent industry (zero otherwise), the proportion of secondary shares issued in the total number of shares offered, a dummy variable equal to one if the offer is an auction (zero otherwise), a dummy variable equal to one if the offer uses an open-price procedure (zero otherwise), a dummy variable equal to one if the offer has previously been postponed (zero otherwise), the number of calendar days from the offer date to the day of the first equilibrium market price (in log). Regressions include market fixed effects and fixed year effects. ***, ***, and * indicate a significativity at the 1%, 5% and 10% level, respectively. Robust t-stats (using White’s heteroskedasticity-consistent standard errors) appear in parentheses. Dependent Variable: Initial Return Stock market return Log (initial offer size) Log (firm’s age) Offer price adjustment Dummy nascent industry Secondary / total shares issued Dummy auction procedure Dummy open-price offer Dummy postponed offer Log (1 + date first market price – date IPO) Log (1 + no. of offers from 2 weeks before to 1 week after) Log (1 + no. of offers from 3 weeks before to 1 week after) Log (1 + no. of offers from 4 weeks before to 1 week after) Log (1 + no. of offers from 5 weeks before to 1 week after) Log (1 + no. of offers from 6 weeks before to 1 week after) Log (1 + no. of offers from 7 weeks before to 1 week after) Log (1 + no. of offers from 8 weeks before to 1 week after) Log (1 + no. of offers from 2 weeks before to 2 weeks after) Log (1 + no. of offers from 3 weeks before to 2 weeks after) Constant

Market fixed effects Fixed year effects Adjusted R² Proba. F-test (all coef. null) Number of observations

(1) 0.610*** (4.19) 3.328* (1.76) -3.757** (-2.26) 1.117*** (7.04) 7.453*** (3.37) 0.010 (0.35) -5.704* (-1.94) -8.056* (-1.70) -8.791** (-2.33) 9.392*** (3.89) -6.007*** (-2.63)

(2) 0.616*** (4.21) 3.356* (1.77) -3.763** (-2.28) 1.104*** (6.88) 7.242*** (3.28) 0.012 (0.40) -5.543* (-1.88) -8.187* (-1.74) -7.545** (-2.05) 9.498*** (3.92)

(3) 0.617*** (4.24) 3.322* (1.76) -3.908** (-2.34) 1.102*** (6.87) 7.118*** (3.24) 0.013 (0.46) -5.390* (-1.81) -8.282* (-1.77) -7.234** (-1.98) 9.525*** (3.95)

(4) 0.623*** (4.22) 3.295* (1.74) -3.893** (-2.33) 1.100*** (6.91) 6.926*** (3.14) 0.010 (0.35) -5.783** (-1.97) -8.721* (-1.86) -7.363** (-1.99) 9.407*** (3.92)

(5) 0.637*** (4.23) 3.264* (1.73) -3.910** (-2.33) 1.097*** (6.92) 6.756*** (3.07) 0.012 (0.40) -6.072** (-2.09) -8.805* (-1.88) -7.612** (-2.06) 9.202*** (3.87)

(6) 0.655*** (4.19) 3.185* (1.69) -3.821** (-2.30) 1.098*** (6.95) 6.783*** (3.08) 0.013 (0.44) -6.315** (-2.20) -8.692* (-1.85) -7.897** (-2.13) 9.128*** (3.86)

(7) 0.672*** (4.22) 3.190* (1.70) -3.788** (-2.30) 1.106*** (7.00) 6.872*** (3.13) 0.017 (0.56) -6.654** (-2.34) -8.631* (-1.85) -7.727** (-2.06) 9.030*** (3.85)

(8) 0.622*** (4.13) 3.295* (1.74) -3.742** (-2.24) 1.119*** (7.10) 7.547*** (3.41) 0.008 (0.27) -5.629* (-1.90) -8.110* (-1.70) -9.117** (-2.42) 9.270*** (3.84)

(9) 0.629*** (4.15) 3.345* (1.75) -3.747** (-2.25) 1.110*** (6.97) 7.363*** (3.34) 0.009 (0.31) -5.561* (-1.88) -8.200* (-1.72) -8.138** (-2.21) 9.331*** (3.86)

-6.927*** (-2.62) -7.870*** (-2.84) -7.574*** (-2.67) -7.740*** (-2.59) -7.383** (-2.44) -8.506*** (-2.64) -4.888** (-2.25)

-47.491 (-1.56)

-47.185 (-1.55)

-45.803 (-1.52)

-44.434 (-1.47)

-43.821 (-1.45)

-42.854 (-1.42)

-42.066 (-1.41)

-47.244 (-1.55)

-5.847** (-2.29) -47.386 (-1.54)

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.263 <0.001 433

0.268 <0.001 433

0.274 <0.001 433

0.272 <0.001 433

0.272 <0.001 433

0.270 <0.001 433

0.276 <0.001 433

0.257 <0.001 433

0.260 <0.001 433

38

Table 3: Descriptive Statistics and Determinants of the Offer Price Adjustment Table 3 reports descriptive statistics of the offer price adjustment from the middle of the price range announced in the prospectus to the final selling price (Panel A) and Probit, Tobit, and Logit estimates of the determinants of the offer price adjustment (Panel B). Auctioned offers (for which no price range is available), postponed IPOs, and offers with an offer price set outside the price range, are excluded from the sample. Explanatory variables are the buy-and-hold return of the SBF 250 index over the 3 months preceding the offer, the initial size of the offer (log of constant euros), the age of the firm at the IPO date (log of number of years), a dummy variable equal to one if the firm is from a nascent industry (zero otherwise), a dummy variable equal to one if the firm is taken public with an open-price offer procedure (zero otherwise), and the number of contemporaneous offers over different periods (in log). Regressions include market fixed effects and fixed year effects. ***, ***, and * indicate a significativity at the 1%, 5% and 10% level, respectively. Robust z-stats appear in parentheses. Panel A: Offer Price Adjustment Mean Median 2.62% 4.10%

Offer price adjustment Number of IPOs that verify

OP < PL 3

OP = PL 38

Std. Dev. 5.72%

Min -24.33%

Max 16.75%

PL < OP < PH 115

OP = PH 140

OP > PH 2

Panel B: Determinants of the Offer Price Adjustment

Stock market return Log (initial offer size) Log (firm’s age) Dummy nascent industry Dummy open-price offer Log (1 + no. of offers in the 3 preceding weeks) Log (1 + no. of offers in the 6 preceding weeks) Constant

Market fixed effects Fixed year effects McFadden R² Proba. LR-stat (all coefficients null) Number of observations

Proba (Offer Price = High Limit of Price Range) Probit Model (1) (2) 0.039*** 0.040*** (3.98) (4.04) 0.374*** 0.383*** (3.26) (3.31) 0.000 -0.001 (0.00) (-0.01) 0.376** 0.367** (2.10) (2.06) -0.605* -0.624* (-1.68) (-1.73) -0.095 (-0.79) -0.167 (-1.23) -7.616*** -7.602*** (-3.76) (-3.75) Yes Yes

Yes Yes

0.169 <0.001

0.171 <0.001

293

293

39

Dependent Variable: Offer Price Adjustment Tobit Model (3) (4) 0.136*** 0.140*** (3.93) (4.07) 0.183 0.156 (0.48) (0.41) 0.213 0.197 (0.65) (0.60) 1.302** 1.240* (2.01) (1.89) -1.471 -1.513 (-1.26) (-1.30) -0.696 (-1.60) -0.478 (-1.00) -2.796 -2.381 (-0.41) (-0.35)

Logit Model (5) (6) 0.134*** 0.139*** (4.06) (4.32) 0.409 0.369 (1.01) (0.92) 0.046 0.049 (0.14) (0.14) 0.937 1.853 (1.43) (1.29) -1.655 -1.665 (-1.26) (-1.26) -0.839* (-1.86) -0.615 (-1.19) -4.794 -4.265 (-0.66) (-0.59)

Yes Yes

Yes Yes

Yes Yes

Yes Yes

293

293

293

293

Table 4: Effects of the Number of Offers before and after the Setting of the Price Range on the IPO Performance Measured from the Preliminary Offer Price (Midpoint of the Price Range) to the First Equilibrium Post-Issue Closing Price Table 4 presents OLS estimates of the determinants of the returns of IPOs measured from the middle of the preliminary offer price range to the first closing (equilibrium) market price. Auctioned offers (for which no price range is available) and postponed IPOs are excluded from the sample. Explanatory variables are the buy-and-hold return of the SBF 250 index over the 3 months preceding the offer, the initial size of the offer (log of constant euros), the age of the firm at the IPO date (log of number of years), the adjustment of the offer price from the middle of the price range, a dummy variable equal to one if the firm is from a nascent industry (zero otherwise), a dummy variable equal to one if the firm is taken public with an open-price offer procedure (zero otherwise), the number of calendar days from the day the offer price is set to the day of the first equilibrium market price (in log), the number of offers completed over the three or in the six weeks preceding the setting of the price range (in log), the number of offers completed from the day the price range is set to the day of the first equilibrium market price (in log), the number of offers completed over different periods preceding the setting of the price range (in log), and the number of offers completed from the setting of the price range to the setting of offer price (in log). Regressions include market fixed effects and fixed year effects. ***, ***, and * indicate a significativity at the 1%, 5% and 10% level, respectively. Robust t-stats (using White’s heteroskedasticity-consistent standard errors) appear in parentheses.

Stock market return Log (initial offer size) Log (firm’s age) Offer price adjustment Dummy nascent industry Dummy open-price offer Log (1 + date first market price – date offer price) Log (1 + no. of offers from 3 weeks before to 1 week after the IPO) Log (1 + no. of offers from 6 weeks before to 1 week after the IPO) Log (1 + no. of offers in the 3 weeks preceding the price range announcement) Log (1 + no. of offers from 3 weeks to 1 week before the price range announcement) Log (1 + no. of offers between the price range announcement and the first market price) Constant

Market fixed effects Fixed year effects Adjusted R² Proba. F-test (all coefficients null) Number of observations

(1) 0.835*** (3.78) 5.119** (2.09) -4.931** (-2.26) 2.496*** (8.60) 10.750*** (3.34) -0.899 (-0.10) 13.723*** (3.77) -9.243*** (-2.65)

Dependent Variable: Return from the Middle of the Preliminary Offer Price to the First Equilibrium Closing Market Price (2) (3) (4) (5) 0.854*** 0.826*** 0.857*** 0.848*** (3.85) (3.80) (3.76) (3.87) 5.078** 4.461* 5.228** 5.139** (2.07) (1.85) (2.11) (2.09) -5.162** -5.553** -4.983** -5.208** (-2.35) (-2.44) (-2.25) (-2.34) 2.515*** 2.516*** 2.514*** 2.505*** (8.76) (8.91) (8.72) (8.68) 10.052*** 9.848*** 10.829*** 10.238*** (3.17) (3.10) (3.38) (3.23) -1.695 -1.805 -2.137 -2.104 (-0.19) (-0.19) (-0.24) (-0.24) 13.500*** 13.604*** 14.563*** 14.394*** (3.76) (3.76) (3.86) (3.90)

(6) 0.871*** (3.82) 5.234** (2.11) -5.105** (-2.29) 2.521*** (8.90) 10.240*** (3.23) -2.382 (-0.27) 14.263*** (3.84)

-10.935*** (-2.64) -9.169*** (-2.69)

-7.416** (-2.35)

-5.209* (-1.75) -79.645* (-1.95)

-4.737 (-1.59) -6.527** (-2.13) -81.691** (-1.98)

-81.111** (-1.98)

-72.928* (-1.82)

-68.543* (-1.71)

-7.981** (-2.39) -83.977** (-2.03)

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.379 <0.001 298

0.385 <0.001 298

0.382 <0.001 298

0.374 <0.001 298

0.387 <0.001 298

0.377 <0.001 298

40

Table 5: Effects of Contemporaneous Offers on Initial Returns Depending on Investment Bank Experience Table 5 reports the number of firms taken public by the 6 most active lead underwriters over the period 1996-2006 and the corresponding total IPO proceeds (Panel A), descriptive statistics of the number of IPOs by the 50 other lead underwriters (Panel B), results of the tests of equality of means and medians of the initial returns, of the stock market returns, of the number of contemporaneous offers (defined as the number of offers completed from four weeks before each offer to one week after), and of the issuer’s age for the two groups of IPOs (high versus low experience underwriters) (Panel C), and OLS estimates of the determinants of the initial returns for the whole sample (columns (1) and (2)), for firms taken public by experienced underwriters (columns (3) and (4)) and for those taken public by poorly experienced underwriters (columns (5) and (6)) (Panel D). Explanatory variables are the buy-and-hold return of the SBF 250 index over the 3 months preceding the offer, the initial size of the offer (log of constant euros), the age of the firm at the IPO date (log of number of years), a dummy variable equal to one if the firm is from a nascent industry (zero otherwise), a dummy variable equal to one if the offer is an auction (zero otherwise), a dummy variable equal to one if the firm is taken public with an open-price offer procedure (zero otherwise), a dummy variable equal to one if the offer has previously been postponed (zero otherwise), the number of calendar days from the offer day to the day of the first equilibrium market price (in log), the number of contemporaneous offers over different periods (in log), and the interaction between contemporaneous IPO volume and a dummy variable equal to one if the firm is taken public by a low experience bank. Regressions include market fixed effects and fixed year effects. ***, ***, and * indicate a significativity at the 1%, 5% and 10% level, respectively. Robust t-stats (using White’s heteroskedasticity-consistent standard errors) appear in parentheses. Panel A: Number of IPOs and Total Proceeds of the Six Top Lead Underwriters Underwriter

Auction Method

Other Method 22 56 21 34 35 42

Total Number of IPOs 24 62 25 56 36 47

Total Proceeds (millions euros) 838 1,275 321 543 373 1,272

BNP Paribas Crédit Agricole Crédit Industriel et Commercial Natixis Oddo Société Générale

2 6 4 22 1 5

Total

40

210

250

4,622

Panel B: Number of IPOs of the 50 Other Lead Underwriters Total number of IPOs 183

Mean 6.54

Median 6

Std. Dev. 3.89

Min 1

Max 14

Panel C: Tests of Equality of Means and Medians (p-values in parentheses)

Initial return Stock market return No. of contemporaneous IPOs Dummy nascent IPOs Firm’s age at IPO

Equality of Means Low Versus High Experienced Underwriters 13.64 / 15.47 (0.51) 5.32 / 6.93 (0.08) 8.68 / 9.95 (0.09) 0.44 / 0.43 (0.85) 16.22 / 17.35 (0.58)

41

Equality of Medians Low Versus High Experienced Underwriters 4.76 / 7.38 (0.11) 6.62 / 7.53 (0.09) 7 / 8 (0.17) 10 / 11 (0.67)

Table 5: Effects of Contemporaneous Offers on Initial Returns Depending on Investment Bank Experience (continued) Panel D: Determinants of Initial Returns and Underwriter Experience Dependent Variable: Initial Return Whole Sample High Experience Low Experience Underwriters Underwriters (1) (2) (3) (4) (5) (6) Stock market return 0.610*** 0.630*** 0.553*** 0.562*** 0.772*** 0.815*** (4.25) (4.26) (3.62) (3.66) (2.76) (2.79) Log (initial offer size) 3.212* 3.165* 2.200 2.087 4.350 4.515 (1.76) (1.73) (0.86) (0.82) (1.37) (1.42) Log (firm’s age) -3.651** -3.647** -1.369 -1.321 -6.208** -6.191** (-2.29) (-2.29) (-0.66) (-0.64) (-2.27) (-2.29) Offer price adjustment 1.117*** 1.113*** 1.206*** 1.203*** 0.760*** 0.731*** (6.92) (6.96) (5.84) (5.86) (2.87) (2.79) Dummy nascent industry 6.942*** 6.588*** 9.898*** 9.724*** 3.138 2.470 (3.13) (2.96) (3.28) (3.22) (0.89) (0.70) Dummy auction method -5.465* -6.270** -5.975 -6.313 -4.938 -6.687 (-1.94) (-2.27) (-1.56) (-1.65) (-1.12) (-1.56) Dummy open-price offer -7.678 -8.233* -7.901 -7.988 -8.107 -9.119 (-1.63) (-1.77) (-1.37) (-1.37) (-1.14) (-1.32) Dummy postponed offer -6.071 -6.305* -6.134 -6.362 -10.342 -10.612* (-1.63) (-1.67) (-1.24) (-1.27) (-1.64) (-1.67) Log (1 + date first market price – date IPO) 9.359*** 8.989*** 10.004*** 9.930*** 10.030** 9.371** (3.94) (3.86) (3.42) (3.38) (2.03) (1.99) Log (1 + no. of offers from 4 weeks before to 1 -5.787*** -3.616* -13.617** week after the IPO date) (-2.96) (-1.93) (-2.37) Log (1 + no. of offers from 6 weeks before to 1 -5.551*** -3.153* -14.855** week after the IPO date) (-2.76) (-1.71) (-2.26) Dummy low experience bank 10.270 13.151 (1.07) (1.13) Log (1 + no. of offers from 4 weeks before to 1 -5.289 week after) × dummy low experience bank (-1.27) Log (1 + no. of offers from 6 weeks before to 1 -5.974 week after) × dummy low experience bank (-1.28) Constant -53.861 -54.798 -29.902 -28.677 -43.967 -43.647 (-1.61) (-1.59) (-0.78) (-0.75) (-0.90) (-0.90) Market fixed effects Fixed year effects Adjusted R² Proba. F-test (all coefficients null) Number of observations

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.277 <0.001 433

0.276 <0.001 433

0.247 <0.001 250

0.246 <0.001 250

0.282 <0.001 183

0.287 <0.001 183

42

Table 6: Initial Returns, Retail Investors’ Demand, and Contemporaneous IPO Volume Table 6 reports descriptive statistics of the percentage of shares reserved to the market procedures (the open-price offer OPO, and the fixedprice offer, FPO), of the proportion of orders exclusively placed by retail investors in the volume of shares allocated to these market procedures through the A orders (the retail investor’s oversubscription ratios), and of the proportion of A orders in the total number of shares offered (Panel A); presents OLS estimates of the determinants of the oversubscription ratios of retail investors, by distinguishing IPOs taken public by poorly experienced banks from offers completed by highly experienced banks (Panel B); and shows estimates of the effects of the retail investors’ oversubscription ratios on initial returns (Panel C). Explanatory variables are the buy-and-hold return of the SBF 250 index over the 3 months preceding the offer, the initial size of the offer (log of constant euros), the age of the firm at the IPO date (log of number of years), the adjustment of the offer price from the middle of the price range to the final selling price, a dummy variable equal to one if the firm is from a nascent industry (zero otherwise), a dummy variable equal to one if the offer has previously been postponed (zero otherwise), a dummy variable equal to one if the firm is taken public with an open-price offer procedure (zero otherwise), the number of contemporaneous offers over different periods (in log), and the oversubscription ratio of the retail investors. Regressions include market fixed effects and fixed year effects. ***, ***, and * indicate a significativity at the 1%, 5% and 10% level, respectively. Robust t-stats (using White’s heteroskedasticity-consistent standard errors) appear in parentheses.

Panel A: Descriptive Statistics of Retail Investors’ Oversubscription Ratios Mean Median Std. Dev. Percentage of shares reserved to the OPO/FPO A orders / shares allocated (oversubscription ratio) A orders / total shares initially offered

21.10% 6.75 1.57

15.00% 2.50 0.44

22.00% 10.09 1.55

Min

Max

5.00% 1 0.002

100.00% 100.00 25.74

Panel B: Determinants of Oversubscription Ratios

Stock market return Log (initial offer size) Log (firm’s age) Offer price adjustment Dummy nascent industry Dummy postponed offer Dummy open-price offer Log (1 + no. of offers from 4 weeks before to 1 week after IPO) Log (1 + no. of offers from 6 weeks before to 1 week after IPO) Constant Market fixed effects Fixed year effects Adjusted R² Proba. F-test (all coefficients null) Number of observations

43

Dependent Variable: Retail Investors’ Oversubscription Ratio Low Experience Bank High Experience Bank (1) (2) (3) (4) 0.311** 0.306** 0.369*** 0.369*** (2.43) (2.37) (3.85) (3.85) 0.447 0.484 -0.996 -0.959 (0.47) (0.50) (-1.04) (-1.06) -1.846** -1.840** 1.278 1.241 (-2.43) (-2.29) (1.52) (1.51) 0.272*** 0.272*** 0.217* 0.219* (2.82) (2.77) (1.94) (1.94) 2.404* 2.353* 4.344 4.291 (1.72) (1.68) (1.41) (1.40) -0.069 -0.462 -0.512 -0.459 (-0.04) (-0.26) (-0.28) (-0.26) -1.423 -1.489 -8.341 -8.417 (-0.72) (-0.76) (-1.14) (-1.14) -3.992** -0.535 (-2.24) (-0.60) -3.539** -0.769 (-2.29) (-0.91) 4.297 4.552 24.967 24.882 (0.28) (0.29) (1.41) (1.44) Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.340 <0.001 124

0.323 <0.001 124

0.151 <0.001 162

0.153 <0.001 162

Table 6: Initial Returns, Retail Investors’ Demand, and Contemporaneous IPO Volume (continued)

Stock market return Log (initial offer size) Log (firm’s age) Offer price adjustment Dummy nascent industry Dummy postponed offer Dummy open-price offer Log (1 + date first market price – date IPO) Log (1 + no. of offers from 4 weeks before to 1 week after IPO) Log (1 + no. of offers from 6 weeks before to 1 week after IPO) Retail investors’ oversubscription ratio Constant

Market fixed effects Fixed year effects Adjusted R² Proba. F-test (all coeff. null) Number of observations

Panel C: Initial Returns, Oversubscription Ratios, and IPO Volume Dependent Variable: Initial Return Low Experience Bank High Experience Bank (1) (2) (3) (4) (5) (6) (7) (8) (9) 1.014** 1.022** 0.170 0.258 0.286 0.876*** 0.890*** 0.499* 0.474* (2.12) (2.13) (0.49) (0.68) (0.75) (2.96) (3.02) (1.81) (1.79) 6.236 6.137 5.109* 4.790* 4.615 2.471 2.304 2.827 3.683 (1.50) (1.50) (1.73) (1.68) (1.66) (0.65) (0.61) (0.81) (1.00) -7.072** -6.458** -2.765 -2.201 -1.642 -2.600 -2.604 -3.604 -4.030 (-2.18) (-2.06) (-1.14) (-1.00) (-0.77) (-0.93) (-0.93) (-1.43) (-1.57) 0.895** 0.878** 0.153 0.168 0.158 1.184*** 1.179*** 0.917*** 0.957*** (2.17) (2.15) (0.49) (0.54) (0.50) (4.00) (4.01) (3.44) (3.49) 7.471 6.987 0.880 0.958 0.676 10.924** 10.661** 6.296 6.381* (1.46) (1.38) (0.23) (0.25) (0.17) (2.45) (2.40) (1.63) (1.67) -2.493 -2.299 -5.458 -2.199 -0.958 -4.483 -4.755 -5.038 -3.721 (-0.31) (-0.28) (-1.12) (-0.43) (-0.18) (-0.72) (-0.75) (-0.87) (-0.66) -6.935 -7.053 -4.360 -4.082 -4.023 -2.711 -2.869 7.101 6.532 (-0.60) (-0.63) (-0.40) (-0.39) (-0.39) (-0.35) (-0.37) (0.97) (0.91) 12.791 12.287 7.778 7.870 7.596 12.927*** 12.923*** 12.023*** 11.640*** (1.49) (1.50) (1.23) (1.22) (1.22) (2.74) (2.74) (3.06) (2.94) -16.969** -6.659 -4.317 -3.822 (-2.01) (-1.12) (-1.49) (-1.49) -18.795** -9.852 -3.879 (-2.02) (-1.37) (-1.39) 2.751*** 2.607*** 2.579*** 1.108*** 1.102*** (5.83) (6.35) (6.65) (2.81) (2.81) -83.036 -71.976 -97.986* -84.925* -74.586 -38.083 -35.824 -59.379 -65.502 (-1.18) (-1.07) (-1.83) (-1.68) (-1.54) (-0.69) (-0.66) (-1.14) (-1.23)

(10) 0.489* (1.84) 3.470 (0.95) -3.992 (-1.56) 0.950*** (3.48) 6.172 (1.62) -4.059 (-0.71) 6.434 (0.89) 11.674*** (2.94)

-3.132 (-1.27) 1.100*** (2.79) -63.077 (-1.19)

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.269 <0.001 124

0.278 <0.001 124

0.543 <0.001 124

0.549 <0.001 124

0.560 <0.001 124

0.222 <0.001 162

0.220 <0.001 162

0.358 <0.001 162

0.360 <0.001 162

0.358 <0.001 162

44