Use of the Proceeds and Long-term Performance of French SEO Firms

Transcription

1 European Financial Management, Vol. 11, No. 1, 2005, Use of the Proceeds and Long-term Performance of French SEO Firms Pierre Jeanneret Institut de l Entreprise, University of Neuchaˆtel and RMF Investment Management pierre.jeanneret@rmf.ch Abstract This paper examines the long-term stock performance of French SEO with rights by looking at the intended use of the proceeds. Firms that raise equity for pure capital structure motives are separated from the ones that use the SEO proceeds to finance specific investment projects. Issuers in the first category are concerned about preserving their financial flexibility and they are expected to evolve in a capital structure irrelevancy framework. On the other hand, issuers in the second category are more inclined to be sensitive to adverse selection problems or agency conflicts and thus, they should be more exposed to under-reaction on the long-run. According to a matching firm methodology, Financing New Investment issuers underperform their benchmark at a rate of 4% to 8% per year over a 36-month horizon while Capital Structure issuers do not show any abnormal performance. These results are robust according to alternative Beta pricing models. In addition, managers of both issuer s types time the SEO after a period of positive abnormal performance in order to sell overpriced securities. However, only the Financing New Investment sample experiences a performance reversal; the abnormal returns decreasing gradually from the issue on, to become significantly negative 24 months after the event. Keywords: seasoned equity offerings; long-term event study; abnormal performance; under-reaction; use of the proceeds; window of opportunity. JEL classification: G14, G32 1.Introduction The long-run stock performance after a seasoned equity offering has been widely documented, especially in the USA. Using a matching firm procedure, Loughran and The author acknowledges financial support from the Swiss National Fund for Scientific Research grant n and from the Kraft Jacobs Suchard Fonds de la recherche dans le secteur de l e conomie. I wish to thank Jean-Franc ois Bacmann, Guido Bolliger, Michel Dubois, Jean-Franc ois Gajewski and an anonymous referee for their helpful comments. I also thank the participants at the 1999 Association Française de Finance meeting in Paris, the 2000 European Financial Management Association meeting in Athens and the 2003 ESA research seminar in Grenoble. # Blackwell Publishing Ltd. 2005, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

2 100 Pierre Jeanneret Ritter (1995) and Spiess and Affleck-Graves (1995) find that issuers underperform their benchmarks in the average proportion of 7% per year over a 5-year horizon. However, it has been shown that time aggregation of abnormal returns (see Conrad and Kaul, 1993; Barber and Lyon, 1997; Kothari and Warner, 1997), pricing model specification (see Jegadeesh, 2000; Mitchell and Stafford, 2000), and time dependence of the events (see Eckbo et al. 2000) have a strong influence on the results. In fact, this literature indicates that we are unable to clearly identify and measure the long-run stock price reaction to a specific event; see Fama (1998). Actually, a large part of the average post-seo performance seems to be due to small firms; see among others, Brav et al. (2000) on the US market and Stehle et al. (2000) on the German market. Loughran and Ritter (1995) and Spiess and Affleck-Graves (1995) attribute the underperformance to the informational asymmetry between managers and investors that allows the firms to issue overvalued securities. This argument, known as the window of opportunity, has been emphasised by Clark et al. (2004). They study the stock sales made by insiders on the secondary market. As the proceeds do not go to the firm itself, free cash flowproblems are eliminated and the abnormal performance can only be imputed to the stock price overvaluation. According to Teoh et al. (1998), managers have the ability to create the window of opportunity by manipulating intermediate corporate earnings before an offering. Afterwards, the stock market temporarily overvalues the firms (see Rangan, 1998) and the price correction is made over a 3 5-year period. Moreover, financial analysts tend to be systematically over-optimistic in their earning forecasts and in the potential growth of the SEO firms. If the analyst is affiliated with the issuer, his predictions are the most optimistic and the post offering underperformance is more severe; see Dechow et al. (1999). However, these findings are mitigated by Hansen and Sarin (1999) who show that over-optimism is common to all high growth firms and is not specific to the equity issuers. Aside the mispricing problems, some researchers underline the impact of the issuing conditions and more generally of the institutional setting on the long-run stock performance. First, the issuance process differs from one market to another and is not neutral to the post-seo performance. In most of the European countries with a legal system based on civil law, the offering with rights remains the predominant issuance process. The subscription rights are issued to prevent the existing shareholders from the dilution of their capital and control. Consequently, they should reduce the informational asymmetry between managers and investors. One could reasonably expect that the long-run stock price reaction after an offering with rights to be insignificant as it is the case in Japan and in Switzerland; see Kang et al. (1999) and Dubois and Jeanneret (2001). In Germany, Stehle et al. (2000) find a 36-month underperformance of 9% that becomes insignificant ( 3.17%) when the abnormal returns are value-weighted. However, in South Africa, the offerings with rights exhibit a long-run underperformance of 40% after 48 months, which is consistent with the results of cash offerings in the USA (see Affleck-Graves and Page, 1996). The aim of the study is to analyse the long-run stock performance of seasoned equity issuers on the French market. It intends to contribute to a better comprehension of the new issue puzzle raised by Loughran and Ritter (1995) in three different ways. First, it extends the research to one of the most important European stock market in terms of market capitalisation. The French institutional setting differs from the US context with respect to the flotation method. This paper focuses on the offerings with subscription rights, which are predominant in France. A wide range of models is employed to determine the long-run

3 Long-term Performance of French SEO Firms 101 abnormal performance. I apply the classic time aggregation of abnormal returns according to an industry class, size and book-to-market matching procedure. Equally- and valueweighting schemes are examined in order to control for a potential size effect, in the sense of Brav et al. (2000) and Stehle et al. (2000). In addition, more sophisticated beta pricing models are employed to check the robustness of the results and the relevancy of the control variables. The second contribution comes from the introduction of the use of the SEO proceeds as a discriminating variable to measure the abnormal performance (the use of the proceeds argument). The use of the proceeds serves to differentiate an equity issue realised in order to finance a specific investment project from an operation made to improve the capital structure and to preserve the different financing sources available to the firm. Both explanations are addressed in the corporate finance theory. Equity financing is often considered as the least efficient financing source in terms of cost of capital, firm reputation, under- or over-investment and free cash flow; see Jensen and Meckling (1976), Myers (1984), Myers and Majluf (1984), Miller and Rock (1985) or Jensen (1986) among others. These models predict that raising newequity conveys a bad signal to the market about the current valuation of the firm or that it gives the opportunity to the managers to use the proceeds inefficiently. However, they exclusively explain the stock price reaction at the announcement of the offering. For the negative impact on the firm value to last in the longrun, the market must under-react to the signal. Behavioural models link the announcement effect with the long-run underperformance of SEO firms. The persistence of the anomaly is explained by investors conservatism (Barberis et al., 1998), overconfidence and self-attribution associated with a high degree of information asymmetry (Daniel et al., 1998) and the slowdiffusion process of information to the market price setters (Hong and Stein, 1999). Another branch of the capital structure literature argues that capital structure choices are irrelevant to the firm value; see Modigliani and Miller (1958) and Miller (1977). In that case, announcements of marginal changes or adjustments in the capital structure are expected to have no impact on the firm value. SEOs made for pure capital structure concerns are less informative than those implying the financing of a specific investment project. Furthermore, if the event implies no market reaction at the announcement, there are no reasons for the under-reaction phenomenon to appear on the long run. Thus, Capital Structure issuers should not experience any long-term abnormal performance. In a survey examining the managers considerations about capital structure decisions, Graham and Harvey (2001) underline the great concern about financial flexibility. A majority of managers declare relying on external financing sources when their internal financial capacity has not been exhausted yet. This observation contradicts the pecking order theory. Managers are inclined to break the hierarchy of financing sources in order not to forego profitable investment opportunities in the future. They use external financing to save internal funds for less favourable periods. Debt represents the main alternative to retained earnings but, once in a while, equity must be issued to reduce the relative leverage and to preserve the debt capacity as well. Firms in that category may be identified by their intended use of the SEO proceeds. They announce that they raise newequity to strengthen the part of equity in the capital structure, to repay debt or to preserve the overall financing capacity so that they may undertake every profitable opportunity in the future. Generally, the SEO proceeds serve to finance going concern investments. Cornett et al. (1998) analyse the bank industry. They differentiate voluntary equity issues from involuntary ones realised in order to respect the minimum required capital standards. The distinction between both types of operations could enter the financing vs.

4 102 Pierre Jeanneret capital structure classification. Cornett et al. examine both announcement returns and long-term stock performance. They find that the abnormal performance is restricted to voluntary equity issuers. These issuers experience a negative announcement effect (twoday CAR ¼ 1.62%) and they underperform their benchmark over a three-year horizon (36-month BHAR ¼ 14.44%). On the other hand, no anomaly can be detected for banks that make involuntary offerings either at the announcement ( 0.39% insignificant) or on the long run ( 0.31%). In addition, the difference in performance between both samples is significant at the announcement as well as on the long-run. These results deserve two observations. First, they support the assumption that long-term abnormal performance due to under-reaction occurs only after a significant announcement effect. Indeed, Cornett et al. (1998, p. 2149) argue that the significant announcement effect of voluntary offerings is consistent with the fact that these offerings contain more information than involuntary issues. The potential under-reaction to a less informative event is then low. Second, the difference in performance between both samples is supportive of the use of the proceeds argument. Pure capital structure concern offerings such as banks involuntary offerings do not experience any long-term abnormal performance, the longrun underperformance being restricted to voluntary offerings. The third contribution of the paper consists in examining the relation between the timing of the SEO and the use of the proceeds argument. The presence of the window of opportunity is tested for both types of offerings. For that purpose, I investigate the stock performance over a 48-month horizon starting one year before the issue and ending 36 months after. If managers take advantage of their superior information to time the equity issue, the abnormal performance over the pre-seo period is expected to be significantly positive. Even if the use of the proceeds allows the differentiation of poor post-event performers from good ones, managers of both types could time the offerings. In that case, Financing NewInvestment issuers would be the only ones to experience a long-term reversal in their abnormal performance. This would be consistent with the hypothesis that behaviourist theories apply only for the most informative events. The remainder of the paper is as follows: in section 2, I present the French institutional setting, the characteristics of French equity issuers and the sample formation. The methodological aspects concerning the measurement of the abnormal performance are exposed in section 3. Section 4 contains the empirical results and section 5 concludes the paper. 2.Seasoned Equity Offerings on the French Market 2.1. The French institutional setting French corporate lawgives the existing shareholders the right to subscribe with priority to a seasoned equity offering. Since 1973, French firms have had the opportunity to raise equity without subscription rights. However, the simplified procedures to waive the rights introduced in 1983 and revised in 1991 has allowed this issuance process to develop. Consequently, the firms can either favour the existing shareholders with an offering with rights or enlarge their financing basis and accelerate the issuing procedure by waiving the subscription rights. The Commission des Ope rations de Bourse (COB), the French security exchange commission, still encourages the firms to use the rights issuance process. Actually, as the stock price fluctuates during the subscription period, the rights prevent the existing shareholders wealth

5 Long-term Performance of French SEO Firms 103 to be altered. Though, if the Shareholders Extraordinary Meeting decides to waive the subscription rights, the COB sees that the existing shareholders interests are preserved. An offering without rights must always respect two principles: First, the issuing price has to be fixed in order not to decrease significantly the value of the firm. The issuing price cannot be inferior to an average price computed on ten consecutive days during a 20-day period before the issue. Second, the issuing conditions must not prevent any existing shareholder to take part to the offering. Until the mid 1990s, a 20-day legal subscription period was reserved for the existing shareholders; in August 1994 the legal period dropped to ten days. This subscription priority cannot be sold as a right. The rigidity in setting the issuing price induces firms to select units, a bundle of stocks and warrants, instead of common stocks alone. Since 1986, French firms have the opportunity to issue units. The optional part of the unit makes it more difficult to value or, in other words, easier to underprice. Therefore, the association of units with cash offerings becomes particularly attractive only from 1991 on; see Chollet and Ginglinger (2001) for an extensive discussion of French units offerings. These legal features have as a consequence that the rights offerings of common stocks are the predominant equity raising operations during the 1980s and the 1990s. At the issuing date, investors have to pay at least 25% of the face value and the entirety of the issuing premium. The rest has to be settled in one or several payments within a fiveyear period after the first quotation. Over the period, the average fraction of the face value relative to the total proceeds represents 28%. It means that at least 80% of the equity raised must be paid at the issue time. Furthermore, Ginglinger (1991, p. 184) argues that it is very uncommon for French listed companies to use this not fully paid method. Therefore, it is assumed to have a negligible impact on the post-seo stock performance. When the Extraordinary General Meeting votes an offering with rights, the Board of Directors must publish an issuing report that includes, among others, the amount raised, the issuing price, the type of shares, the issuing conditions and the use of the proceeds. The legal constraint of announcing the use of the SEO proceeds implies that investors are able to better control the managers actions. If the proceeds are not spent as intended, the firm could be penalised in its future attempts to call for external financing. Moreover, the information is common to all issuers, which prevents the sample to be biased Data and samples Between 1984 and 1998, the Socie te des Bourses Françaises (SBF the French Clearing House) reported in its annual reports (L Anne e Boursie`re) 402 seasoned equity offerings with rights. 1 Subsequent offerings that occur within the 36-month period after the issuer s preceding operation are eliminated. They represent 54 observations. The analysis periods of both issues overlap and their abnormal performances 1 The data taken from the SBF have the advantage to include operations that were realised and not intended to be realised, which could be the case with the register files of the COB. The sample of 402 SEO with rights does not include operations combining an equity issue with a capital reduction as stated in the issuing prospectus. It is restricted to issues of common stocks and hence it does not contain offerings of units and investment certificates. Direct cash offerings (44 operations) are not considered because they have specific informational content (see Gajewski and Ginglinger, 2002) and because they are not numerous enough to be examined separately.

6 104 Pierre Jeanneret nb of SEOs Financing New Investment Capital Structure Fig. 1. Rights offerings with stocks made in France between 1984 and The total sample includes 232 rights offerings of common stocks 118 Capital Structure offerings and 114 Financing NewInvestment issues. The use of the proceeds is taken from the issuance prospectuses. are embedded, which would influence the interpretation of the results. Finally, issuers must have data available one year prior to the offerings about their stock prices and their balance sheets (to compute the book-to-market ratios). Moreover, the issuing prospectus 2 must be available to identify the use of the SEO proceeds. Because these data are unavailable at the COB, the SBF and at the firm itself, 65 additional operations are taken out of the sample. Yet, the final sample includes 232 rights offerings of common stocks that are illustrated in Figure 1. From the final sample, I form two sub-samples according the use of the proceeds: 118 Capital Structure and 114 Financing NewInvestment offerings. Descriptive statistics of the sub-samples are presented in Table 1. To be ranked in the Financing New Investments sample, the issuer must specify that the SEO proceeds will be spent in internal growth projects (a newplant, a newproduct or a newmarket segment) or in external growth opportunities (acquisitions). Firms in the Capital Structure sample explicitly declare that their main motives to issue equity are improving their capital structure, preserving their financial flexibility or repaying debt. The Panel A of Table 1 presents data about market capitalisation, book-to-market ratio, the offering size and the issuing discount. The first four columns showmeans and medians of both sub-samples while the last column exhibits their difference in medians. The issuer size is measured by the pre-seo market capitalisation of the firm. For both samples the distribution of the size is highly skewed to the right, indicating that small firms are more numerous than large ones. This finding justifies the use of both equally and value weighting schemes in the computation of the abnormal performance. The book-to-market ratios are calculated by dividing the book value of equity at the end of the last fiscal year by the pre-seo market value. The book values are collected in the annual edition of the Dictionnaire Dafsa Desfosse s des Socie te s, that contains the balance sheet and the profit and loss statements of French listed companies. The offering relative size is given by the rawseo proceeds divided 2 I thank Gre goire Henriotte at the COB and Marc Douëzi at the SBF for their help in collecting the issuing reports.

7 Long-term Performance of French SEO Firms 105 Table 1 Descriptive statistics of the French SEO with rights, Firm size is the market value at the beginning of the SEO month taken from Datastream. Book-to-market ratio (BM ratio) is computed as the book value of equity at the end of the pre-seo fiscal year divided by the market value of equity at the beginning of the SEO month. The offering relative size is measured by the rawoffering proceeds divided by the market value at the beginning of the SEO month. The issuing discount is given as the difference between the first post- SEO market price and the issuing price divided by the market price. Assets, debt, cash and cash floware accounting values. All book values are taken from the Dictionnaire Dafsa Desfossés des Socie tés. The use of the proceeds and the rawproceeds are taken from the issuance prospectuses. The issuing discount is taken from the SBF report L Année Boursière. Wilcoxon rank tests are given in parentheses. Capital Structure Financing NewInvestment Difference 118 observations 114 observations CS FNI Mean Median Mean Median Median Panel A: Issuers general characteristics Issuer size (millions FRF) ( 1.56) BM ratio (0.34) Offering relative size (1.46) Issuing discount ( 0.47) Panel B: Issuers financial structure data Debt-to-assets ratio a (3.01) Cash-to-assets ratio a ( 3.06) Cash flow-to-assets ratio a a significant at the 1% level, b significant at the 5% level ( 4.58) by the pre-seo market capitalisation. The issuing discount corresponds to the difference between the first post-seo market price and the issuing price, divided by the market price. The median characteristics of both samples are not statistically different at convenient levels. According to that, the average Capital Structure issuer is similar to the average Financing NewInvestment issuer. In Panel B of Table 1, data about issuers financial structure are exposed. The debt-toassets ratio, the cash-to-assets ratio and the cash flow-to-assets ratio are computed, at the end of the pre-seo fiscal year, with accounting data taken from the Dictionnaire Dafsa Desfosse s des Socie te s. Capital Structure issuers appear to be significantly more levered than Financing NewInvestment firms. They also have less cash and less cash flow. These findings generate three remarks. First, the equity issue for the Capital Structure sample is expected to lower the leverage ratio, to improve cash and to compensate for a lower cash flow. Consequently, it is consistent with either the pecking order and the financial flexibility argument. Second, on average, Financing NewInvestment issuers select an

8 106 Pierre Jeanneret equity offering while other financing sources seem to be available to them. The voluntary SEO argument of Cornett et al. (1998) and the window of opportunity argument of Loughran and Ritter (1995) could apply. Third, because they differ in leverage, Capital Structure and Financing NewInvestment issuers could have different equity systematic risk coefficient. Abnormal performance measures should take this possibility into account. Figure 2 illustrates the repartition of the final sample across industry classes. The industry classification is based on the Worldscope one-digit SIC codes. Then, I separate transport from communication and internet, banks and insurance companies from real estates and from other financial firms, and services to the community (utilities, medical care facilities or laboratories and educational services) from the other private services. I obtain twelve classes. The repartition of the sample across industries is not different at the 5% level ( 2 ¼ with 11 degrees of freedom) from the one of the entire population. It is hence representative of the French market. 3.Benchmark Selection and Abnormal Performance Measure 3.1. Standard benchmarks The use of a matching firm that is not subject to the event reduces the biases; see Barber and Lyon (1997) and Kothari and Warner (1997). The selection of the matching firm is based on three criteria: (a) it has not undertaken an equity issue Other financial companies 8% [8%] Services to the community 2% [1%] Private services 14% [13%] Natural resources and construction 6% [3%] Food 4% [6%] Transformation industries 11% [12%] Real estate 7% [7%] Bank and insurance 12% [10%] Retail 11% [9%] Transport 4% [3%] Manufacturing industries 20% [22%] Communication and internet 1% [5%] Fig. 2. Repartition of the SEOs across the industry classes. The industry classification is based on the first-digit SIC codes, with the largest classes being sub-divided into more precise categories. The 12 final industries: Natural resources and construction, Food, Transformation industries, Manufacturing industries, Transport, Communication and internet, Retail, Bank and insurance, Real estate, Other financial companies, Private services, Services to the community. Percentages in brackets are the expected proportions according to the French economy.

9 within the last 36 months; (b) it belongs to the same industry class; 3 and (c) it minimises the following distance: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 size i;0 size c;0 d i;0 ¼ þ BM 2 i;0 BM c;0 2 size where d i,0 is the distance measure of SEO firm i and matching firm c at event month 0, size is the market value at the beginning of month 0, BM is the book-to-market ratio at the beginning of month 0, 2 size ; is the variance of the series of the variable size (BM). 2 BM Long-term Performance of French SEO Firms 107 Each control variable is standardised in order to prevent one variable from having more weight then another. Size is measured by the pre-seo market value of the firm. The book-to-market ratio is determined in the manner explained in section 2.2, using the pre-seo market value in order to take into account the current firm valuation. If during the post-seo analysis horizon a matching firm engages itself into a SEO process or is delisted, it is switched to the next closest firm at the current month on a point forward basis. 2 BM ð1þ 3.2. Determining the abnormal performance The long-run abnormal performance is computed over a 36-month horizon after the issue. Monthly returns and market values for the period are taken from the Association Franc aise de Finance (AFFI) database which includes more stocks than Datastream over this period. Data on monthly stock prices, dividends and market prices for the period are from Datastream. The returns are computed cum-dividend. The series of stock prices and dividends are adjusted for capital structure changes. First, I compute the average monthly adjusted returns (AR t ) up to month 36 after the issue, month 1 being the issue month: AR t ¼ Xnt i¼1 w i R SEO;it R c;it where R SEO,it is the return on the seasoned equity offering firm i in event month t R c,it is the return on the benchmark over the same period w i is the weight of the firm i in the sample 4 n t is the number of seasoned equity offering firms. ð2þ 3 Restricting the set of potential matching firms within the issuer s industry class allows industry effects on the stock returns to be taken into account. 4 For the equally-weighted average the weight is 1/n t. For the value-weighted average, the weight of firm i in the sample is computed in a two-stage method. First, in order to avoid scale problems related to non-synchronous events, the market value. of the SEO firms is scaled by the total market capitalisation at the SEO month: av i;mi ¼ mv i;mi nm j¼1 mv j;m i, where av i, m is the adjusted market value of the firm i at the issuing date m, mv i, m is the market value of firm i at date m and n m is the number of stocks present in the database at date m. Second, the weight is calculated as the proportion of the SEO firm adjusted value relative to the sum of all the adjusted values in the sample: w i ¼ av i;mi np. k¼1 av k;m i

10 108 Pierre Jeanneret The negative abnormal performance could be caused by small firms so that the global economic impact is overwhelmed; see Brav et al. (2000) and Stehle et al. (2000). Therefore, in order to control for this phenomenon, the abnormal performance is determined according to both an equally and a value-weighted methodology. The cumulative average monthly adjusted return for months 1 to T (T ¼ 6, 12, 24 and 36 months) is computed as: CAR T ¼ XT t¼1 AR t I also compute the average buy and hold abnormal return (BHAR): " # Y BHAR T ¼ XnT T Y T w i 1 þ R SEO;it 1 þ R c;it where n T is the number of firms at month T. The null hypothesis to be tested is the following: i¼1 t¼1 t¼1 ð3þ ð4þ H 0 : CAR T (BHAR T ) ¼ 0 vs:h 1 : CAR T (BHAR T ) 6¼ 0: Under the null, the standardised cumulated average abnormal returns adjusted for cross-sectional variance and first order autocovariance is distributed as Student-t; see Ritter (1991). Empirically, the buy and hold series are highly skewed. Consequently, I use the skewness adjusted t-statistic proposed in Barber et al. (1999). A (nonparametric) sign test based on the proportion of positive abnormal returns (CAR i,t or BHAR i,t ) is presented as well to check the results Alternative models to determine the abnormal performance Three different beta pricing models are used to check the robustness of the results. The abnormal performance is measured by the Jensen s alpha. Performance measured with the standard unconditional CAPM (U-CAPM) r p;t ¼ U p þ p r m;t ð5þ where r p,t is the monthly return on a portfolio p in excess of the risk free rate, 5 and r m,t is the monthly return on the SBF market index 6 in excess of the risk free rate. The constant term in the regression ð U p Þ is the Jensen s alpha of the portfolio p according to the model used (here U stands for the U-CAPM). 5 The risk free rate is given by the 1-month euro FRF. 6 The SBF index is a value-weighted index adjusted for dividend and capital structure operations that contains the firms with the highest market value. The number of firms included in the index increases with the number of companies listed.

11 Long-term Performance of French SEO Firms 109 Performance measured with the Fama and French three-factor model (FF 3-factor) r p;t ¼ FF p þ p;1 r m;t þ p;2 SMB t þ p;3 HML t ð6þ where SMB t is the difference in return between small firms 7 (bottom half) and large firms (top half) at month t, HML t is the difference in return between high (top 30%) and lowbook-to-market firms (bottom 30%) at month t. Both SMB and HML factors can be contaminated by the firms involved in a SEO; see Loughran and Ritter (2000) and Brav et al. (2000) for a justification. Therefore, the factors are purged of those companies. The same incidence affects the market risk premium. Though, Loughran and Ritter suggest to use a non-purged market index. They base their argument on the fact that the CAPM is an equilibrium model and consequently, the market index containing all the listed companies is the only one appropriate to measure the risk premium. One can object to this rationale that the CAPM is developed under the hypothesis of a perfect market. Moreover, the long-run abnormal stock performance after an event affecting the firm is considered as an anomaly. In that sense, not integrating the firms that cause the anomaly into the benchmark represents a better approximation of perfect market conditions. As the aim of the study is not to focus on the epistemology of long-run abnormal performance determination, the Jensen s alphas are estimated separately with a non-purged and a purged market index. Performance measured with a conditional CAPM (C-CAPM) Eckbo and Smith (1998) and Eckbo et al. (2000) showthat the abnormal performance is better measured by a conditional Jensen s alpha when the expected returns are time varying. Thus, I use a conditional CAPM with a time varying beta as defined in Ferson and Schadt (1996, p. 430, equation 4): r p;t ¼ C1 p þ d 0; r m;t þ d 0 z t 1 rm;t ð7þ where d 0, is the unconditional beta, z t 1 ¼ Z t 1 E(Z) is the vector of the set of centred information variables (instruments), d is the vector of coefficients with equal dimension of Z t 1. I examine the explanatory power of the instruments used by Bossaerts 8 and Hillion (1999) as well as other standard instruments employed in previous studies. I retain the three most significant variables as instruments: 9 the long-term bond yield 7 In order to avoid the extreme returns due to the very lowmarket price of companies in financial distress, firms in lowest quintile according to the market capitalisation are not taken into account during the factor formation. 8 I am grateful to Peter Bossaerts for giving me access to his data on the French market. 9 As the choice of relevant instruments is not the main preoccupation of this study, I do not present the results. However, they are available upon request.

12 110 Pierre Jeanneret (Datastream s France bond yield government 10 years), the price to earnings (MSCI), the inverse of the market price (1 over the SBF market index monthly price). Construction of the calendar-time portfolio returns (r p ) Following Eckbo et al. (2000), I form equally- and value-weighted calendar-time portfolios of issuers. A separate portfolio is created for each use of the proceeds category. Starting with the initial issuer in the sample, the SEO firms enter the portfolio at their offering month and they are held during 36 months (unless they are delisted before the 36th month). The portfolio is rebalanced each time a firm is included or removed (or delisted), using current value weights or equal weights. The same procedure is applied to form the portfolios of matching firms with weights identical to those of the issuers portfolios. In order to test the use of the proceeds argument, a zero-investment portfolio is built where the Financing New Investments issuers are sold short to finance the long position in the Capital Structure issuers. The same procedure is applied to both matching firms portfolios. The Jensen alpha is expected to be negative for the first zero-investment portfolio and not significantly different from zero for the second one. One drawback of the matching firm methodology is that it does not specifically control for differences in equilibrium returns between event and matching firms. Eckbo et al. (2000) attribute the underperformance to the incomplete adjustment for risk with standard benchmark methodologies. Zero-investment portfolios long in the event firms and short in the matching firms controls for such risk adjustment problems. If their alphas are significantly different from zero, the abnormal performance cannot be fully explained by differences in systematic risk. 4.The Long-run Stock Performance of SEO with Rights 4.1. The impact of the use of the proceeds on the long-run performance The results for the long-run performance according to the matching firm methodology are presented in Table 2. Cumulative abnormal returns are shown on the left part of the table and buy and hold abnormal returns on the right. Each panel entails the results of the equally-weighting scheme, those of the value-weighting scheme and the non-parametric performance. The abnormal performance of the whole sample (all uses of the proceeds included) is shown in Panel A. The equally-weighted abnormal returns are negative and significant over the 24-month horizon ( 15.38% for the CAR and 18.62% for the BHAR). After 36 months, the abnormal returns are no longer significant at a conventional level. When the performance is value-weighted, the CARs lose their significance at every horizon while the significance of the BHARs persists up to 24 months after the issue. However, the magnitude of the abnormal returns is lower ( 11.31% for the 24-month BHAR and 4.63% for the 36-month BHAR), showing that the underperformance is sensitive to the size of the issuer, but the change is not sufficient enough to drastically alter the results. In that sense, the results of French rights offerings are different from what is observed in Germany; see Stehle et al. (2000). On the long-run, median CAR is significantly negative ( 12.98% after 36 months) but median BHAR is not. All in all, the results of the whole sample are unclear about a persistent underperformance after SEO with

13 CAR are computed as: Table 2 Long-run performance based on a matching firm methodology. CAR T ¼ XT X nt t¼1 i¼1 w it R SEO;it R c;it where T is for 6, 12, 24 and 36 months, R SEO, it is the return on SEO firm I at month t, R c, it is the return on the matching firm of I at month t, w it is the weight of I in the sample at month t and n t is the number of firms in the sample at month t. The t-stat for the CAR is computed as in Ritter (1991) as: p ffiffiffiffiffi t stat T ¼ CAR T n T=½T var þ 2 ðt 1ÞcovŠ 1=2 where T is the horizon considered (number of months), var is the average (over 36 months) crosssectional variance and cov is the first-order autocovariance of " the AR t series. The average BHAR# is calculated as X nt Y T Y T w it 1 þ R SEO;it 1 þ R c;it i¼1 t¼1 where R SEO, it is the return on the SEO firm I at month t, R c, it is the return on the matching firm of I at month t, T is the holding period considered (6, 12, 24 or 36 months), w it is the weight of I in the sample at month t and n T is the number of the SEO stocks for the T-month period. The t stat is the skewness adjusted t statistic of Barber et al. (1999) and is calculated as: t stat T ¼ ts T þð1=3þ ts 2 T = p ffiffiffiffiffi pffiffiffiffiffi n T skewt þð1=6n T Þ n T skewt where ts T is the standard t stat value computed at month T, skew T is the skewness of the BHAR series at month T and n T is the number of SEO firms at month T. The sign test is computed as: ðp T 0:5Þ=ð0:5 ð1 0:5ÞÞ 0:5 p ffiffiffiffiffi n T where pt is the percentage of positive abnormal returns at month T and n T is the number of observations at month T. Cumulative Abnormal Returns t¼1 Buy and Hold Abnormal Returns 6-month 12-month 24-month 36-month 6-month 12-month 24-month 36-month Panel A: Rights offerings of common stocks (232 observations) Equally weighted performance Mean (%) 8.05 b 9.27 b b a b t stat ( 2.49) ( 2.02) ( 2.33) ( 1.57) ( 2.91) ( 1.75) ( 2.36) ( 1.72) Value weighted performance Mean (%) a 8.23 b b 4.63 t stat ( 1.51) ( 1.48) ( 1.30) ( 0.59) ( 3.09) ( 2.34) ( 2.04) ( 0.52) Long-term Performance of French SEO Firms 111

14 Cumulative Abnormal Returns Table 2 Continued. Buy and Hold Abnormal Returns 6-month 12-month 24-month 36-month 6-month 12-month 24-month 36-month Non-parametric performance Median (%) 9.55 a a 7.46 a Sign test ( 3.18) ( 1.86) ( 1.28) ( 2.82) ( 2.95) ( 1.21) ( 1.85) ( 1.66) Panel B: Rights offerings of common stocks, Capital Structure (118 observations) Equally weighted performance Mean (%) t stat ( 1.20) ( 0.86) ( 1.73) ( 0.19) ( 1.50) ( 0.86) ( 1.10) ( 0.41) Value weighted performance Mean (%) t stat ( 0.84) ( 0.77) (0.21) (0.35) ( 1.86) ( 1.50) (0.42) (0.63) Non-parametric performance Median (%) Sign test ( 1.79) ( 1.13) ( 0.48) ( 1.18) ( 1.82) ( 1.15) ( 0.30) ( 0.20) Panel C: Rights offerings of common stocks, Financing New Investment (114 observations) Equally weighted performance Mean (%) b b b b b b b t stat ( 2.46) ( 2.11) ( 2.04) ( 2.11) ( 2.58) ( 1.62) ( 2.27) ( 2.16) Value weighted performance Mean (%) b b 9.59 a b b b t stat ( 1.47) ( 1.37) ( 2.02) ( 1.97) ( 3.15) ( 2.24) ( 2.57) ( 2.04) Non-parametric performance Median (%) 9.61 a a a 8.61 b b b Sign test ( 2.81) ( 1.60) ( 3.32) ( 2.91) ( 2.35) ( 0.57) ( 2.29) ( 2.12) 112 Pierre Jeanneret a significant at 1%, b significant at 5%,

15 Long-term Performance of French SEO Firms 113 rights. They only give partial support to the underperformance pattern of US equity issuers; see Loughran and Ritter (1995), Spiess and Affleck-Graves (1995) or Jegadeesh (2000). Panels B and C present the results of respectively the Capital Structure and Financing NewInvestments samples. As expected, Capital Structure issuers exhibit no abnormal performance. Equally-weighted CAR and BHAR over the 36-month horizon are negative but not different from zero ( 2.20% for CAR and 6.63% for BHAR). The value-weighted performance is negative on the shorter horizons and turns to become positive after 36 months without being significant. Median CAR and BHAR are not significant either at a convenient level. These results support the hypothesis that an equity issue made to improve the capital structure does not affect the firm market value. The long-run performance of the Financing NewInvestments sample is significantly negative at a 5% level. The 36-month cumulative abnormal returns range from 20.64% to 28.95% depending on the weighting scheme and on the computation method. The medians are also significantly negative, 16.11% for the CAR and to 12.30% for the BHAR. The underperformance is significant in most of the horizons less than 36 months, the 12-month horizon showing the least cases of significance (only the 12-month value-weighted BHAR is significant). Hence, Financing New Investments issuers are poorly performing firms from the time they issue newequity and up to three years after the event. On one hand, firms caring about financial flexibility, that preserve their financing sources, do not underperform their benchmark over a 36-month horizon. On the other hand, firms that are financially constrained to issue equity to undertake a specific investment project are penalised on the long-run. They underperform their benchmark by 7.4% and up to 10.5% per year. Based on the use of the proceeds argument, I compare the 36-month means and medians The hypothesis is the following: H 0 : R FIN;36 > R CS;36 vs: H 1 : R FIN;36 R CS;36 where R s;36 is the mean or median abnormal return (CAR or BHAR) of sample, s ¼ {CS, FNI}, at the 36-month horizon. As it can be seen in Table 3, the H 0 hypothesis is rejected in five of the six cases. Only the median BHAR of the Capital Structure sample is not significantly greater than its Financing New Investment correspondent. Otherwise, for all weighting schemes and abnormal returns aggregation methods, both issuer types are different relative to their long-term stock performance. These findings emphasise the use of the proceeds argument. 10 The underperformance of the Financing NewInvestment issuers could be due to short sale restrictions. At the time of the issue, these restrictions do not be permit prices to reflect 10 As pointed out by the referee, one reason for the abnormal performance of Capital Structure issuers not to be significant could be that the standard deviation of their abnormal returns is greater than that of Financing NewInvestment issuers. Cross-sectional variances of both subsamples are not found to be significantly different, given any analysis horizon, weighting scheme or time aggregation method. Consequently, the non-significance of the Capital Structure sample cannot be attributed to a larger dispersion in its abnormal returns. Results are not presented here for the sake of brevity.

16 114 Pierre Jeanneret Table 3 The difference in long-run abnormal performance according to the use of the proceeds. CS denotes the Capital Structure issuers and FNI represents the Financing NewInvestment issuers. The CARs and BHARs are taken at the 36-month horizon. The tested hypotheses are: H 0 : R FIN;36 > R CS;36 vs H 1 : R FIN;36 R CS;36, where R s;36 is the mean (median) abnormal return (CAR or BHAR) of sample, s ¼ {CS, FNI}, at the 36-month horizon. The t-stat for the difference in means (given in parentheses) is computed as: t-stat ¼ R CS;36 R FNI;36 r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; 2 R CS;36 n CS þ 2 RFNI;36 n FNI where 2 s;36 is the cross-sectional variance of the 36-month abnormal returns (CAR or BHAR) of sample, s ¼ {CS, FNI}, and n s is the number of observations in sample, s. The test of the difference in medians (given in parentheses) is a Wilcoxon rank tests. 36-month CAR 36-month BHAR CS FNI CS-FNI CS FNI CS-FNI Mean equally-weighted (%) b (1.70) Mean value-weighted (%) a (2.88) Median (%) b (1.71) b (1.74) b (1.74) (1.01) a significant at 1%, b significant at 5% the negative views of the pessimists. 11 During the sample period, the First segment of the French market is a forward market ( Re` glement Mensuel, RM) where short sales are not restricted. When splitting the Financing NewInvestment sample between RM issuers (50 operations) and issuers listed on other segments (64 operations), no differences in abnormal returns are noticed. The 36-month average CAR of RM issuers equals 22.33% while other issuers exhibit a 24.11% CAR (the t-stat of the difference equals 0.10). Results are not sensitive to the computation method and to the analysis horizon. This finding shows that the long-run underperformance is not due to short sale restrictions Is the abnormal performance consistent with alternative benchmarks? Does the underperformance found with the classical methodology survive the test of other benchmarks? Because the decontamination of the market index does not affect the significance of the results, only Jensen s alphas estimated with a non-purged market index are shown in Table 4. The portfolios of issuers and matching firms are presented in the first four columns and the alternative models in rows. Panel A contains the results of the equally-weighted portfolios and Panel B presents those of the value-weighted portfolios. The first general comment is that the results of alternative benchmarks support the matching firm results. Beta pricing models also detect a significant underperformance for 11 I thank the referee for pointing out this explanation.

17 Table 4 Abnormal performance (Jensen s alpha) computed from alternative models. The mean r p is the mean monthly (equal- or value-weighted) return on the portfolio p in excess of the monthly risk free rate (1-month euro FRF). Jensen s alpha is estimated with three alternative models. Unconditional CAPM (U-CAPM): r pt ¼ U p þ pr mt where r mt is the monthly return on the SBF market index in excess of the risk free rate at month t, Fama and French (1993) three-factor (FF 3-factor): r pt ¼ FF p þ 1;p r mt þ 2;p SMB t þ 3;p HML t where SMB t is the difference in monthly return between small and large firms (size is measured by equity market value), HML t is the difference in monthly return between high and low book-to-market firms, SMB and HML are purged from SEO firms conditional CAPM (C-CAPM): r pt ¼ C p þ d 0;p þ d 1;p y t 1 þ d 2;p pe t 1 þ d 3;p mi t 1 rmt where y t 1 is the one month lagged French government 10-year bond yield, pe t 1 is the market aggregated one month lagged price to earnings ratio, mi t 1 is the one month lagged inverse of the SBF market index price. The coefficients are estimated using OLS. The numbers in parentheses are t statistics. CS Match CS FNI Match FNI CS-Match FNI-Match CS-FNI Match CS-FNI Panel A: Equally-weighted portfolios Mean rp rf (%) Std. deviation (%) U-CAPM alpha (%) b b 0.84 b 0.01 (1.11) (0.39) ( 2.03) (0.30) (0.72) ( 2.00) (2.47) (0.03) Long-term Performance of French SEO Firms 115

18 Table 4 Continued. CS Match CS FNI Match FNI CS-Match FNI-Match CS-FNI Match CS-FNI FF 3-factor alpha (%) a b 1.13 b 0.14 (1.05) ( 1.22) ( 2.57) ( 1.09) (1.84) ( 1.97) (2.36) (0.39) C-CAPM alpha (%) b b 0.96 a 0.01 (1.32) (0.47) ( 2.06) (0.44) (0.87) ( 1.97) (2.63) ( 0.04) Panel B: Value-weighted portfolios Mean rp rf (%) Std. deviation (%) U-CAPM alpha (%) a (0.47) (0.16) ( 1.93) (1.53) (0.15) ( 2.58) (1.72) ( 1.01) FF 3-factor alpha (%) a b 1.56 a 0.74 (1.78) ( 1.00) ( 2.76) (0.61) (1.91) ( 2.14) (3.15) ( 1.25) C-CAPM alpha (%) b a 0.99 b 0.67 (0.82) (0.21) ( 2.07) (1.76) (0.32) ( 2.89) (2.10) ( 1.11) a significant at 1% b significant at 5% 116 Pierre Jeanneret

19 Long-term Performance of French SEO Firms 117 the Financing NewInvestments portfolio while other calendar-time portfolios do not exhibit significant alphas, at the 5% level. These findings contradict those of Brav et al. (2000) and Eckbo et al. (2000) showing that the calendar-time approach eliminates the underperformance. However, they are in accordance with Jegadeesh (2000). The results also underline the importance of the use of the proceeds as a variable to differentiate the issuers. The significance of the alphas, when it is the case, persists across the alternative models. The introduction of supplementary risk factors emphasises the significance of the Financing NewInvestments portfolio results. Another interesting feature is the closeness of the alphas given by both conditional and unconditional CAPM. Consequently, one can reasonably say that the abnormal performance is not time varying. The alphas of the Capital Structure portfolio (1st column of Table 4) range between 0.37% and 0.46% per month in the equally-weighting scheme and between 0.16% and 0.61% per month in the value-weighting one. They are never significant at the 5% level. The alphas of the Financing NewInvestments portfolio (3rd column of Table 4) are significantly negative according to all models except for the unconditional CAPM in the valueweighting scheme. They range from 0.47% to 0.95%. The alphas of both matching portfolios are never statistically significant (columns 2 and 4 of Table 4), which outlines the fact that the abnormal performance cannot be attributed to non-event firms. Zero-investment portfolios long in issuers and short in matching firms (columns 5 and 6 of Table 4) exhibit different alphas. For the Capital Structure minus Matching Firm portfolio, the regression intercepts are not significant while those of the Financing NewInvestments minus Matching Firms are significantly negative for all models. After controlling for possible differences according to market systematic risk, 12 Financing NewInvestments issuers still underperform their matching firms, underlining the impact of the event on the long-term stock performance. The last two columns of Table 4 present the results of zero investment portfolios long in Capital Structure issuers (matching firms) and short in Financing NewInvestment issuers (matching firms). The portfolio of issuers shows significantly positive alphas according to all models. Firms raising equity for pure capital structure concerns outperform, on the long-run, the ones that need external cash to finance specific investment projects. Note that the difference between the portfolios of matching firms is not different from zero. This corroborates the fact that the anomaly is restricted to the Financing NewInvestments issuers The timing of SEOs and the abnormal performance In this sub-section, I extend the analysis of the long-run performance to a 48-month horizon starting one year prior to the issue month. The aim is to examine if the use of the proceeds argument remains valid. In other words, is the long-run underperformance of Financing NewInvestments issuers preceded by a period of over-performance? Does the difference between the two main groups remain when the analysis horizon includes the pre-seo period? The 12 Not shown in the table for the sake of brevity, the coefficients for the market factor are indeed significantly different from zero. Both issuers and matching firms are not expected to have similar equilibrium returns. However, the underperformance observed for Financing New Investments issuers is not due to that difference.

20 118 Pierre Jeanneret explanation of the long-run stock underperformance of equity issuers advanced by Loughran and Ritter (1995) is that managers take advantage of their superior information to issue equity when it is over-valued. Then, investors under-react to the announcement of the offering and the underperformance persists on the longrun. To test the window of opportunity argument, I use the matching firm methodology. I select the matching firms at the SEO month according to the same procedure as the one described in section 3. Then, I compute the BHAR starting one year prior to the issue month and ending 36 months after. The results are presented in Table 5. Panel A concerns the Capital Structure sample while Panel B addresses the Financing NewInvestment offerings. The Capital Structure sample exhibits a significant over-performance during the pre-seo period in both weighting schemes (þ15.77% for the equally-weighted BHAR and þ12.94% for the value-weighted BHAR). The median BHAR is also positive but not significant. Over the same horizon, Financing NewInvestment issuers are noticed to outperform their benchmark in proportions similar to the other issuers (equally weighted BHAR ¼þ15.96%; value-weighted BHAR ¼þ13.01%). In fact, the difference in the pre-seo abnormal performance between the two samples is not statistically significant: the equally-weighted difference equals 0.19% (t-stat ¼ 0.02), the value-weighted difference equals 0.07% (t-stat ¼ 0.01), and the Wilcoxon rank test is also insignificant ( 0.38). These results indicate that managers time the equity issue within the window of opportunity independently from the intended use of the proceeds. Over the 48-month horizon, the two samples show different patterns. The positive performance in terms of means and median persists for the Capital Structure issuers all through the post-seo period but without being significant. On the other hand, the abnormal performance of the Financing NewInvestment sample decreases during the post-seo period to become significantly negative at the two longest horizons. The equally-weighted 48-month BHAR is equal to 29.01% whereas the value-weighted abnormal return reaches 25.41%. These figures are lower than their corresponding 36-month post-seo BHARs in Table 2. It means that firms with the highest pre-seo over-performance do not experience the worst post-seo underperformance. The 48-month median BHAR also becomes significantly negative, showing that the number of underperformers increases with the horizon length. The evolution of the long-term abnormal performance around the SEO is somehow puzzling. All firms tend to issue equity within the window of opportunity but only the Financing NewInvestment ones underperform on the long-run. The market is unable to identify the firm type before the issue. The descriptive statistics in Table 1 showthat both issuers types are not different except in their motivations to raise equity. The Capital Structure sample presents an interesting pattern. In fact, the performance increases up to the SEO month and then tends to decrease during the first year after the issue. This decrease is not as steep as the corresponding Financing NewInvestment pattern but it is quite noticeable in the value-weighted performance. This could mean that the market is aware of the managers timing ability. Yet, the overpricing correction is done completely within a month period after the issue. Conversely, the market is unable to correctly incorporate into the stock price the information contained in a Financing NewInvestment offering over a 36-month horizon. The uncertainty implied by the decision to issue equity combined with the difficulty to value the newinvestment project may constitute one cause of the anomaly.