Seasoned Equity Issues in a Closely Held Market: Evidence from France

This paper studies the shareholder wealth effect associated with 278 seasoned equity issues in France from 1986 to 1996. The sample includes public offerings and rights issues of common stock alone and of unit. On average, there is a negative stock price change of 0.58% at the first announcement date of common stock rights issues. There is no significant return at the announcement date of public offerings or unit rights issues. We find strong evidence of negative abnormal returns at the issue date and during the subscription period for common stock issues, whether rights or public. Abnormal returns at the announcement and issue dates are related to the main characteristics of the offerings and especially to the corporate control mechanisms. In France, public offerings are chosen by larger firms and result in a more diffuse ownership. Furthermore, abnormal returns at the announcement of a public offering are not related to company specific risk, but are negatively related to the expected take-up of the current shareholders. In a closely held market such as the French market, public offerings result in a less concentrated ownership, which is better news for investors than rights issues. Conversely, the expected shareholder take-up does not affect abnormal returns in the case of rights issues. In this case, the market reacts less negatively when company specific risk is lower, stock price performance prior to the announcement is weaker and the aim of the issue concerns either an investment project or an acquisition. These results shed new light on the equity financing paradox in a closely held market.


Introduction
Despite the lower flotation costs of rights issues, most American firms choose public offerings to raise capital. This puzzling evidence is often referred to as the equity financing paradox. Several explanations, suggesting that rights offerings are more expensive in other ways, are put forward to resolve this paradox. These explanations include capital gains taxes (Smith, 1977), shareholder selling costs (Hansen, 1988), differences in ownership (Hansen and Pinkerton, 1982) and adverse selection costs (Eckbo and Masulis, 1992). The purpose of this study is to provide new international evidence to explain the choice of a flotation method.
The relative frequencies of rights issues and public offerings differ strongly from one country to another. In the U.S., the overwhelming majority of corporations choose the firm commitment underwriting method. In Japan, according to Kang and Stulz (1996), 66 % of the common stock offerings over the period 1985-1991 were public offerings. In most European countries and in Australia, rights issues are the primary flotation method 1 . In this paper, we examine seasoned equity offerings on the French market, where rights issues represented 77 % of all equity offerings over the period 1986-1996. The French market presents different characteristics that convey interest to this study. First, even if the rights offering is the primary flotation method in France, French firms select the public offering method more frequently than in other European domestic equity markets.
Second, French firms can choose to issue units (stocks with attached warrants) in their seasoned offerings. Third, the French institutional setting for public offerings is different from the American setting. Fourth, the French market is a relatively closely held market.
French public offerings rules differ from American rules in three ways. First, in most cases, shares are first offered to current shareholders, on a pro-rata basis, for ten days on average, but this priority cannot be sold like a right. Second, there is a regulation constraint on the issue price: it cannot be less than the average price of ten consecutive daily stock prices chosen among the twenty daily prices before the issue. This regulation constraint, associated with the priority period, makes the public offering method risky. Third, public offerings are generally underwritten through a standby-underwriting contract. Firm commitments can only be used when there is no priority period.
Furthermore, French firms issue units quite frequently, which include both shares of common stock and warrants, traded separately thereafter. In most cases, warrants can be considered as an incentive to raise interest in new offerings. Another explanation may be provided by the choice of a flotation method: the regulation constraint on the offering price in public offerings is hardly effective as far as units are concerned since the valuation of the warrants depends on volatility estimation.
In this paper, we focus in particular on differences in corporate control that may explain the choice between the different flotation methods. The French equity market, like most European equity markets, is a closely held market. Hansen and Pinkerton (1982) argue that firms with concentrated share ownership will choose rights issues. They document that the average level of control exceeds 61 % of the outstanding common stock of the rights issuers. We find that the percentage of shares held by known ownership in France, on average 64 %, is not significantly different for the rights and the public equity issuers. However, our evidence shows that the expected shareholder take-up is much lower for public offerings than for rights issues: on average 39 % and 55 % respectively. This evidence is consistent with the results of Eckbo and Masulis (1992). They argue that when the degree of current-shareholder participation in the issue is high, firms will prefer rights offerings; when their participation is low, firms are more likely to issue public offerings. In their model, the issuing firms show adverse selection to a degree that, all else being equal, is inversely related to the current shareholder take-up. They predict that the average market reaction to an uninsured rights offer is close to zero, and is less negative in the case of a standby rights offer than in the case of a firm commitment offer.
Our evidence is not consistent with these predictions. We find negative abnormal returns to rights issues and no significant abnormal returns to public offerings. In France, public offerings are chosen by much larger firms than rights issues and often contain an international part of the issue. According to Eckbo and Masulis' data, the average size of American firms does not seem to differ for standby rights and firm commitment issuers. We can reasonably think that, in France, public offerings are accompanied by a lower degree of asymmetric information. Furthermore, abnormal returns are not related to company specific risk in the case of public offerings. But, the market reacts less negatively when current shareholders decide not to subscribe. In a closely held market, public offerings, which result in a less concentrated ownership on average, are better news for investors than rights issues, whether standby or uninsured. Conversely, the expected shareholder take-up does not affect abnormal returns in the case of rights issues. In that case, the market reacts less negatively when company specific risk is lower, stock price run-up prior to the announcement is weaker and the aim of the issue concerns either an investment project or an acquisition.
The paper extends the empirical literature on the equity-financing paradox in the following directions. First we investigate the seasoned equity offerings in France, by using a large database of 278 equity issues. The sample covers the period from January 1, 1986 to December 31, 1996. It includes both rights and public offerings of common stocks and units.
We select a real announcement date while most event studies in France use a legal announcement date. Second, we provide evidence on expected shareholder subscription as a factor of the choice of flotation method. Third, we investigate market reaction during the offering period. We show that the offering period abnormal returns are linked to the type of underwriting contract used.
The paper proceeds as follows: in Section I, the sample and the methodology are described. In Section II, we give evidence on flotation costs. Section III presents evidence on the valuation effect of offerings announcements, including the effect of ownership dispersion on stock returns. Section IV examines the offering period abnormal returns. Summary and conclusions are contained in Section V.

A. Equity issues in France over the period 1986-1996
All equity issues taking place on the French market can be identified through the annual reports of the COB 2 (" Commision des Opérations de Bourse "). The empirical tests are made over this eleven-year period. Prices are extracted from the SBF database (" Société des Bourses Françaises "), and daily returns are calculated after taking into account changes in capital structure. In our initial sample, 590 equity issues from 1986 to 1996 were reported and classified with their characteristics. The final sample excludes all issues that do not meet the following criteria: -The issue involves a single type of security and does not come with a stock dividend; -The issue does not involve a common stock reduction; -The issue does not involve a restructuring plan; -The firm does not publish important information, such as earnings, at the same time as the issue announcement 3 .
These criteria produce a sample of 278 offerings described in table I. Standby rights and uninsured rights are presented in table II. In France, most rights issues are underwritten.
Sometimes, issuers select a partial standby contract in which the underwriter guarantee covers less than 100 % of the issue. Partial standbys occur frequently when current shareholders offer commitments prior to the subscription. We can observe that 11 % of all public offerings are not underwritten. The average size of the equity rights offerings is FF 354 million and the average size of public offerings is FF 891 million. On average, the subscription price is 83 % of the prevailing stock price (78 % for rights issues and 98 % for public offerings). French firms are required to fix the rights subscription price at the agreement date, which is on average 10 days before the beginning of the subscription period. This period lasts on average 15 days. The subscription price in public offerings is disclosed nearer to the beginning of the issue (on average 4 days before), and the subscription period is shorter (8 days long on average). This relatively long period explains why the issuer and the underwriter have to anticipate secondary market prices when determining the offer price. The average number of shares offered, as a percentage of the outstanding and new shares, is 22 % (24 % for rights issues and 16 % for public offerings).

B. Time-line of important dates
The first announcement date considered is the Board Meeting date when the firm decides to issue equity. At this date, insiders may be informed. The first institutional announcement of offering information generally comes in the registration statement filed with the COB. No newspaper in France gives coverage of equity issues, contrary to newspapers in the US such as the " Wall Street Journal ". We however had access to information in the database of the " Européenne des données ", which includes all announcements of " Agence France Presse " and those of the main French financial newspapers. As a matter of a fact, the announcement of the issue by " Agence France Presse " often takes place the day before publication in the financial dailies. Most of the previous studies on securities issues in France analyze the wealth effect at the BALO date, which is the legal announcement date. The present study is the first to use a real announcement date.
The " Européenne des données " date (EDD) corresponds to the first mention of the offerings in the press. It does not often specify all their characteristics, but generally indicates the amount and the period of the issue. The EDD date is taken into account only if it is prior to the COB date. Thus, in our study, the announcement date is the first of the two following dates: date of the " Européenne des données " database in addition to the COB date.
In the case of rights issues, there is a legal period of seven days between the BALO date and the issue date. In the case of public offerings, there is no legal period. The issue date is then frequently the same as the BALO date or even for a few offerings the day before.
The issue date is also considered. When there is a priority period, the subscription period is of ten days at least. Figures 1 and 2 report the time-line of these dates. Table IV reports the characteristics of these different dates in our sample.

C. Event-study methodology
A standard event-study is performed in order to measure the impact on prices of a seasoned equity offering announcement. Define day 0 as the event date. The purpose is to test if mean excess returns are significantly different from 0 around the relative day of announcement.
If R it designates the observed logarithmic return for security i and R mt the index return at date t 4 , The excess return can be calculated by the difference between R it and a reference N it , which corresponds to a normal return in the absence of any event 5 . The parameters 6 are estimated over a period between 220 days and 21 days before the day of announcement ( ) t = 0 . The cross-sectional mean excess return ( ) RAM t is then calculated at each date of the event period that is defined over 40 days around the date of announcement. In order to know if mean excess returns are significantly different from 0, the null hypothesis ( : ) H RAM t 0 0 = is tested. The two main tests are parametrical 7 . The first statistic (1) is obtained by dividing the mean excess return by a time-series standard estimation. 4 The index return used in this study is the SBF (Société des Bourses Françaises) index that is calculated over the 250 most liquid securities. The index return takes into account the reinvestment of dividends. 5 Three reference returns can be defined : -the index return : N R it mt = -the mean return measured over an estimation period before the event period : In the case of the market-adjusted return, the coefficients are estimated by using the OLS values over the estimation period. However, because of non-synchronous trading, OLS estimates are inconsistent. Dimson (1979) has proposed a methodology that takes this bias into account. Scholes and Williams (1977) and Fowler and Rorke (1983) give estimates that take returns autocorrelation into account. These estimates are calculated by considering two lags and two leads.
7 Under normality hypothesis, T tps and T trans are distributed Student with T − 1 and N − 1 degrees of freedom.
However, as returns normality is not proved, the sign test is also used. The sign test compares the number of positive abnormal returns and the number of negative abnormal returns: where N + and N − designate the strictly positive and negative returns. T sig is distributed unit normal. (1) The second statistic (2) is obtained by dividing the mean excess return by a cross-sectional standard estimation. (2) Table V shows average relative costs for the different flotation methods. The French flotation costs seem much lower in relative value than those underlined by Eckbo and Masulis (1995).

II. Evidence on the flotation costs
They show that the direct flotation costs of public underwritten offerings are over 6 % of the issue proceeds, but are only 4 % for standby rights offerings and 2 % for uninsured rights offerings. Our results are however very similar to the figures of Singh (1997) for rights issues.
In France, the part of the offering subscribed by outside investors, which is the only costly part of the issue, is much lower than in the United States, and this may explain the differences in flotation costs.
The results of American studies show that the flotation costs are higher for firm commitments, even after controlling for the issue characteristics (for instance Eckbo and Masulis, 1995).
Underwritten public offerings seem also to be more expensive than underwritten rights issues in France. We compare flotation costs for the two flotation methods after controlling for the issue and firm characteristics. Table VI shows the coefficient estimates in cross sectional regressions on the issue characteristics. The dependent variable is either total or banking flotation costs divided by offer proceeds. The explanatory variables include the type of the issue, the gross proceeds, the systematic risk, the prior long-run performance, the percentage change in outstanding shares due to the offer, the underwritten part, the current shareholder take-up, and one binary variable accounting for the existence of an international part. The results are reported for all issues, public offerings and rights issues, and the tests are based on White's (1980) heteroskedastic consistent variance covariance matrices. From table V, it seems that rights issues direct flotation costs are significantly lower than public offerings costs. The equity financial paradox, which is related to the lower costs of rights offerings in the United States, is therefore confirmed in France.

[Insert table V] [Insert table VI]
The regression intercept is positive and significant, which indicates a fixed component of flotation expenses. The flotation costs decrease with the gross proceeds of the offer, as a result of economies of scale. Higher stock risk increases underwriter fees and therefore flotation costs, but only for rights issues. This result shows that in the case of public offerings, the underwriter does not bear a higher risk if the stock becomes riskier. In the same way, stock price performance prior to the announcement of a rights issue positively affects flotation costs. Firms with better performances may present a greater risk for underwriters during the subscription period.
In France, issues are generally underwritten through a standby contract. In this case, the underwriter has to buy the shares to which the public does not subscribe. Consequently, public offerings imply higher risk than rights offerings, because the public part is greater. But, in the case of public offerings, the flotation costs do not increase when stock risk becomes higher. In these conditions, in the case of public offerings, either the part subscribed to by the underwriter is weak or the underwriter does not sell the acquired shares at a loss. In other words, it seems that public offerings do not result in very negative news for the underwriters.
In the same way, stock price run-up does not affect the flotation costs in the case of public offerings. In these conditions, firms going public do not schedule their issues, or the underwriters do not care about the timing of these issues. The absence of significance for prior performance and systematic risk in the case of public offerings is consistent with a better information content of these issues.
Furthermore, a higher percentage change results in lower costs in the case of a rights issue.
This may result from a higher current shareholder take-up for the offerings with a greater change in the number of shares. The variable public measures the rate of subscription of new investors. It is equal to one if all the current known shareholders 8 subscribe in proportion to their previous number of shares. It is higher than one if some of the current shareholders decide not to subscribe. As the flotation costs mainly come from the part subscribed to by outside investors, the variable public positively affects the costs in the case of a public offering. In the case of a rights issue, the ownership structure is not likely to change.
Consequently, the variable public does not affect flotation costs. 8 The known shareholders are those who are named in the registration prospectus.
Finally, the choice of an underwritten offer increases flotation costs. The international part of the issue involves additional costs 9 .

III. Ownership structure and valuation effect of the equity offering announcement
Stock price reaction to seasoned equity offerings made by American firms has been extensively examined. A large number of studies convincingly show that seasoned public offerings are associated with a decrease in the announcing firm's stock price, especially when the flotation method used is firm commitment: on average -3 % for industrial issuers over the two-day announcement interval (see, for instance, Asquith andMullins, 1986, andMasulis, 1992 10 ). In other countries, less evidence has been found on the valuation effect of equity issues announcements. In Japan, for instance, Kang and Stulz (1996) report a significantly positive reaction of 0.45 % at the announcement of public offerings. In most European countries, empirical studies about issue announcements find positive or insignificantly negative market reactions (see for instance Loderer and Zimmerman, 1987, Bohren, Eckbo and Michalsen, 1997, or Gebhardt and Heiden, 1998. These different findings seem puzzling. We shall try to explain them by focusing on corporate control and shareholder take-up differences, as most European countries are closely held markets. We shall shed light on French issuers ownership and subscription characteristics, and show how they may explain abnormal returns around equity issue announcements. A. Ownership structure and shareholder take-up Hansen and Pinkerton (1982) argue that firms with concentrated share ownership will choose rights issues. They document that the average level of control exceeds 61 % of the outstanding common stock of the rights issuers. From table VII, we find that the percentage of shares held by known ownership in France, on average 64 %, is not significantly different for rights and public equity issuers. Bohren et al. (1997) show that the average percentage of outstanding equity held by the 20 largest shareholders varies from 54 % in the 1980-1984 period to 61 % in the 1985-1993 period. They find that there is little evidence that the average ownership characteristics vary systematically with the flotation method.

[Insert table VII]
In France, the issue prospectuses indicate the percentage of the new shares that will be subscribed to by the main shareholders. We take this variable as an estimation of the shareholder take-up. It may be lower than the real shareholder take-up as the unknown shareholders may subscribe too. But we think that it is a better estimation than the percentage of rights traded on the stock exchange, assuming that a right is traded at most once over the period, which is considered in Bohren et al. (1997). Furthermore, even if our variable underestimates shareholder take-up, it does reflect the informed's take-up, and is thus more relevant in an adverse selection context. We distinguish the shareholder precommitments from the shareholder take-up. In the first case, main shareholders offer their guarantee for all or part of the issue, instead of bank underwriting.
Our evidence shows that the expected shareholder take-up is much lower for public offerings: on average 39 % for public offerings and 55 % for rights issues. This evidence is consistent with the results of Eckbo and Masulis (1992) and Kothare (1997): he shows that rights offerings lead to a more concentrated ownership structure while public issues result in a more diffuse ownership. It means that the current shareholder take-up is high in the first case and low in the second case. According to Bohren et al. (1997), the average shareholder take-up is lower for standbys than for uninsured rights.

B. Valuation effect at the first announcement date (AD)
The event study is completed around two dates (Board Meeting date and the first date between the COB date and the EDD date  Allen and Faulhaber (1989) or Welch (1996), only high quality firms have an interest in selling stock in two stages, because they are confident that their stock price will rise, while low quality firms may prefer to sell all the shares at the current price.
Unit issues may therefore signal undervalued firms, and this may explain the lack of negative reaction on unit issue announcements.
The absence of significant reaction in the case of public offerings at the date of announcement is a notable result, as it is very different from the American results. However, Table IX shows that market reaction is more negative for rights issues than for public offerings, although not significantly so.
11 Only the abnormal returns estimated with Dimson's (1979) method are put forward. The other results based on the market model (OLS estimates), the index return, and the mean lead to similar conclusions. Only the test based on a time-series standard-deviation is shown. The test based on the cross-sectional standard deviation and the non-parametrical test lead to similar results.

[Insert table IX]
Figure 3 provides evidence for a significant announcement effect during the event period in the case of common stock rights offerings.

C. Factors affecting abnormal returns at the announcement date
A cross-sectional analysis is completed in order to explain the width of mean abnormal returns. This analysis makes it possible to observe differences between direct issues and rights issues around the date of announcement. The whole sample is divided into two groups depending on the type of issue (rights or public). The least squares regression models are developed, using hypothesis tests based on consistent estimates of covariance matrices allowing for heteroskedasticity, as in White (1980). The following regression model (3)  , ln , , is the cumulative average abnormal return between relative date t 1 and relative date t 2 . Dps is an indicator value that takes the value of 1 if the offer is a rights issue.
GP designates the gross proceeds. Rap represents the ratio of the number of offered shares to the number of outstanding shares before the offer. Obj is an indicator value that is equal to 1, if the firm issues in order to acquire another firm or to invest in a project. Ai designates company specific risk 12 . More precisely, this volatility is defined by the market-adjusted residual variance of the daily stock abnormal returns for the 200 days of the estimation period prior to the event period. Perf designates the long-run performance prior to the event period. It is calculated by the cumulated stock returns over the 200 days of the estimation period. Pub represents the following ratio: Pub Percent offered to external investors known shareholders = − 100% .
In order to capture the announcement effects and to explain these effects, model 3 is estimated by calculating abnormal returns at the precise date of announcement. The results are presented in table X.

[Insert table X]
The cross-sectional regression yields an adjusted R² of 7 % for the total sample, 6 % for rights issues, and 9 % for public issues 13 . The gross proceeds receive a significantly negative coefficient, for all the samples, indicating a size effect in abnormal returns. The size effect is consistent with a price pressure hypothesis, but also with an information-based explanation.
Presumably, the more overvalued the equity, the larger the incentive to issue more shares.
The variable Perf receives a significant negative coefficient, suggesting that the adverse selection effects are more pronounced when the stock price performance prior to the offering announcement is higher. This variable is significant for the whole sample and the rights issues, but not for the public offerings. Asquith and Mullins (1986)  It is interesting to note that these arguments are not valid for French public offerings. Prior performance, company specific risk and issue purpose have no effect on the price reaction at public offerings announcements, suggesting that there is less adverse selection effect for these issues. This evidence is not consistent with the third hypothesis of Eckbo and Masulis (1992): issuers with a high shareholder take-up, whose stock is undervalued, show smaller preannouncement stock price run-ups than issuers with a low shareholder take-up. This is consistent with Gajewski and Ginglinger (1996). As the flotation costs are lower for rights issues, the overvalued firms always choose the rights issuing flotation method.
Conversely, the undervalued firms choose either the public offering or the rights issue. Their choices depend on current shareholders' subscription. In these conditions, asymmetric information effects exist in the case of a rights issue, but disappear in the case of a public offering.
What explains stock market reaction at the announcement of a public offering is the public participation. First, the percentage change in outstanding shares has a positive effect on abnormal returns: the market reacts more favorably to issues for which the dilution is greater.
The positive value of α 6 indicates that the market reaction is less negative when current known shareholders decide not to subscribe (totally or partially), for the whole sample and the public offerings. These findings are not consistent with the Eckbo and Masulis (1992) model, according to which shareholder take-up should have a positive effect on market reaction. To our knowledge, there has been no evidence that shows a positive relation between shareholder take-up and abnormal returns. In their paper, Eckbo and Masulis explain the market reaction after deduction of flotation costs; they find that the variable share concentration, which should be correlated with shareholder take-up according to these authors, has no effect on prices for the sample of industrial issuers and a negative effect for the sample of utilities issuers. Bohren et al. (1997) show that the market reacts more favourably to issues for which the proportion of insiders (board members and the CEO) is greater, but for which the proportion of common stock held by the 20 largest shareholders is not significant.
Overall, our evidence confirms the Gajewski and Ginglinger (1996) model. Overvalued firms select rights issues; undervalued firms decide to issue rights if the current shareholder take-up is high and public offerings otherwise. Investors can infer from a public issue that the firm is undervalued; this may explain insignificant reaction at the announcement of public offerings.
When a firm issues through rights, investors cannot infer if the firm is over or undervalued: the market reaction is negative. Furthermore, we find that the market reacts less negatively to public offerings, all the more so as the take-up of the current known shareholders is low.
To explain these results, we shall explore corporate control mechanisms. The French market is characterized by closely held firms. Majority ownership is relatively common. As Schleifer and Vishny (1997) suggest, there may be costs of concentrated holdings: large investors represent their own interests and may treat themselves preferentially at the expense of other investors and employees, especially if there is a substantial departure from the one-share-onevote rule. In France, in most firms, there is a one-share-two-votes rule when shares are owned on average for more than 2 to 4 years. Furthermore, as pointed out by Holmström and Tirole (1993), concentrated ownership reduces the monitoring of the firm by stock market participants, thereby reducing the amount of information available about the firm. This, in turn, increases spreads. Kothare (1997) finds that rights issues are associated with an increase in proportionate bid-ask spreads, while public underwritten offerings are followed by a decrease in proportionate bid-ask spreads. He attributes his results to the different effects of the two flotation methods in the firm's ownership structure. Rights issues increase ownership concentration while public offerings decrease it. This may explain why the market reacts less negatively to public offerings, all the more so as the take-up of the current known shareholders is low: ownership becomes more diffuse after public offerings and this is good news for external and minority investors.
These findings shed new light on the adverse selection effect. Overall, our evidence shows that information asymmetry around equity issue announcements is lower for public offerings than for rights issues. It may be possible that on the US market, the current shareholder takeup is favorable news for external investors, because the costs of concentrated ownership are lower than the adverse selection costs. But in a closely held market, this does not appear to be the case and the current shareholder take-up becomes unfavorable news for the market's participants. These findings have to be confirmed in other closely held markets, and especially for NASDAQ small firms for which ownership is relatively concentrated (see Kothare, 1997).

IV. Price reaction during the offering period
To investigate price reaction on the date of completion of the operation, we select the first day of issue and the following twenty days. We can safely assume it is during this period that stockholders decide whether to exercise or to sell their rights in the case of rights offerings.

A. Price reaction at the issue date (ID)
From table XI, the market reaction at the issue date is significantly negative for common stock issues, but not for unit issues. Most studies find insignificant returns at the issue date for public offerings (see Asquith andMullins, 1986 or Eckbo andMasulis, 1992). However, Hansen (1988), Eckbo and Masulis (1992), Kang and Stulz (1996) and Singh (1997) find significant negative returns during the rights offer period.
In an efficient market, prices should take into account all the information available on the day the operation is announced. Price changes on the issue day cannot result from new information. However, according to some authors (Mikkelson andPartch, 1988 or Korajczyck, Lucas andMcDonald, 1992), as the issue can be withdrawn after the announcement, not withdrawing it indicates that the stock remains overvalued despite the negative reaction following the announcement. In France, no operation is cancelled after the registration; consequently, it seems this explanation cannot be accepted.
The other explanations usually provided refer to the current stockholders' and investors' behaviour at the date of completion. According to Lease, Masulis and Page (1991), as investors buy stocks in the primary market, sellers will exceed buyers in the secondary market bringing prices down.

[Insert table XI]
The results presented in exercise or to sell their rights. If they are risk-averse, shareholders who want to sell their rights will do so as soon as possible, as long as rights are in the money. Investors who buy the rights will buy stocks on the primary market rather than on the secondary market, bringing secondary market prices down. Furthermore, standby underwriters, who have to buy all remaining shares at the end of the subscription period, will be net stock sellers during the subscription period. This argument may explain the negative price reaction for rights offerings.
All previous studies show insignificant results for public offerings at the issue date. We find a significant negative abnormal reaction during the public offering subscription period. In the French context, current shareholders have a priority subscription period and public offerings are underwritten through standby contracts and not through firm commitment contracts. We show that the shareholder take-up is low for public offerings. Current shareholders do not subscribe, and outside investors will buy shares in the primary market rather than in the secondary market, bringing the stock prices down. We should bear in mind that current shareholders have a priority subscription period, but when the known shareholder take-up is low, outside investors may also subscribe immediately. This setting may explain the similar results for rights and public issues around the issue date.

B. Factors affecting abnormal returns at the issue date
The issue date does not bring anything new to the investors. In this way, announcement effects are completely incorporated into prices. However, a stock market reaction is observed at the beginning of the operation. Regression model (4)

[Insert table XII]
The cross-sectional regression yields very low R², except for public offerings. The market reaction at the issue date is not significantly different for public offerings and rights issues, after controlling for the issue characteristics. For the rights issues, none of the included variables are significant. For public issues, a size effect is detected. Furthermore, when current known shareholders decide not to subscribe, the market reaction is more negative.
These variables are also significant when we consider the cumulative average abnormal returns for a 10-day period after the issue. This confirms our previous analysis: at the issue date, stock prices decrease all the more so as the current shareholder take-up is low.

V. Conclusion
This paper provides new evidence on the equity flotation method choice. In France, like in most European countries, rights issue is the primary flotation method. Our findings show that the abnormal returns around the announcement dates are significantly negative for rights issues and negative, but not significantly so, for public offerings. Public seasoned issues announce less negative news than rights issues in France, but not significantly so. These   (1987) for Switzerland, Hietala and Löyttyniemi (1991) for Finland and Bohren et al. (1997) for Norway. 2 The COB plays the same role as the SEC in the US.
3 French firms quite often announce equity issues at the same time, or immediately after earnings publications.
This evidence is consistent with Korajczyk, Lucas and McDonald (1991), who find that equity issues follow shortly after earnings publications. 4 The index return used in this study is the SBF (Société des Bourses Françaises) index that is calculated over the 250 most liquid securities. The index return takes into account the reinvestment of dividends.
5 Three reference returns can be defined : -the index return : N R it mt = -the mean return measured over an estimation period before the event period : 6 In the case of the market-adjusted return, the coefficients are estimated by using the OLS values over the estimation period. However, because of non-synchronous trading, OLS estimates are inconsistent. Dimson (1979) has proposed a methodology that takes this bias into account. Scholes and Williams (1977) and Fowler and Rorke (1983) give estimates that take returns autocorrelation into account. These estimates are calculated by considering two lags and two leads.
However, as returns normality is not proved, the sign test is also used. The sign test compares the number of positive abnormal returns and the number of negative abnormal returns: where N + and N − designate the strictly positive and negative returns. T sig is distributed unit normal. 8 The known shareholders are those who are named in the registration prospectus. 9 All the issues with an international part are public offerings. This is why the variable is not included for rights issues.
10 For a survey of market reactions to equity issue announcements, see Smith (1986) or Eckbo and Masulis (1995). 11 Only the abnormal returns estimated with Dimson's (1979) method are put forward. The other results based on the market model (OLS estimates), the index return, and the mean lead to similar conclusions. Only the test based on a time-series standard-deviation is shown. The test based on the cross-sectional standard deviation and the non-parametrical test lead to similar results.
12 Following Dierkens (1991), this variable, calculated by the residual volatility of the equity of the firm, may be a proxy for information asymmetry. But, this variable measures specific risk of the firm, which is not exactly the same as the risk of asymmetric information.
13 Inclusion of a dummy variable for unit issues does not alter the results: this variable showed an insignificant coefficient and was dropped.              , is the cumulative average abnormal return between relative date t 1 and relative date t 2 . Dps is an indicator value that takes the value of 1 if the offer is a rights issue. GP designates the gross proceeds. Ai designates the firm's specific risk. Perf designates the long-run performance of the firm over 200 days prior to the announcement of the issue. Obj is an indicator value that is equal to 1, if the firm issues in order to acquire another firm or to invest in a project. Pub represents the proportion of the issue that is offered to new investors divided by the part of the issue that could be offered to them if all the current known shareholders would subscribe to their parts in the offering. It is defined by: Pub Percent offered to external investors known shareholders = − 100 % . Rap designates the percentage change in outstanding shares due to the offer (new shares divided by old shares). The figures in square brackets represent the statistics of Student. *, **, *** denote significance of the test at the 0.1, 0.05 and 0.01 levels respectively.