OPERATING MEASURES, IPO VALUATION AND THE AFTERMARKET PERFORMANCE: Perspective from internet bubble period

Transcription

1 OPERATING MEASURES, IPO VALUATION AND THE AFTERMARKET PERFORMANCE: Perspecive from inerne bubble period Tarek MILOUD Absrac In his paper, we empirically invesigae European and Unied saes iniial public offerings (IPOs) o provide a comparaive case on he inernaional evidence on he long-run performance of IPOs. Specifically, he paper examines he relaion beween iniial reurns and long-erm performance in he IPO marke. We also examine wheher he choice of a performance measuremen mehodology direcly deermines boh he size and power of saisical es, as documened in previous sudies (Michell and Safford (2000); Loughran and Rier (2000); and Brav, e al. (2000). We use wo samples, he firs one consiss of 277 IPOs realised beween 997 and 999 in he Euro.NM and he second one consiss of 277 paired IPOs realised during he same period in NASDAQ. We use all long erm performance measures and we observe he exisence of long erm abnormal reurns for our wo samples. While, he fads or invesor overreacions and divergence of opinions hypoheses do no apply in explaining he afermarke performance of our IPOs samples. Keywords: Iniial public offerings; Underpricing; Afermarke performance; T. MILOUD is Professor of Finance a he Graduae School of Business a Chambéry, Deparmen of Finance and Accouning, Savoie Technolac; 2, avenue Lac d Annecy; 7338 Le Bourge du Lac Cedex - FRANCE. .miloud@esc-chambery.fr. The auhor appreciaes he helpful commens of M. LEVASSEUR and R. CAUBBOT a Universié Caholique de Louvain (BELGIUM).

2 . INTRODUCTION AND BACKGROUND A large volume of research has demonsraed ha invesors purchasing iniial public offerings (IPO s) of common socks earn a large posiive abnormal reurn in he early afermarke period. However, researchers have documened ha he gains from early price appreciaion are no sufficien o compensae he losses ha occur hroughou subsequen price declines. This aricle focuses on he empirical invesigaion of long-erm performance and survival paerns of European firms ha issued heir iniial public offerings in Euro.NM marke during he period 997 hrough 999. Mos of he previous research in his area has been based on IPOs in U.S. sock marke, which focused on New York Sock Exchange and NASDAQ. These sudies used cumulaive abnormal reurns (CAR) as performance measures of in documening IPO long-erm performance and considered marke index and maching firms, based on marke capialisaion and marke-o-book raion, as benchmarks for evaluaing he relaive performance. The conclusions abou long-erm performance of IPOs have differed considerably across sudies ranging from a poor performance o a somewha neural performance. Rier (99) finds a significan mean marke-adjused reurn of -29.3% a he end of he hird year following he offering for a sample of,526 IPO s over he period from 975 o 984. Furher, Rier (99) repors ha he underperformance is concenraed among younger firms and firms ha wen public in he heavy-volume years. Indeed, for more esablished firms going public, and for hose ha wen public in he ligh-volume years of he mid and lae 970 s, here is no long run underperformance. IPO s ha are no associaed wih venure capial financing, and hose no associaed wih high-qualiy invesmen bankers, also end o do especially poorly. These findings are in conformiy wih Loughran and Rier (995) who, for 4,753 U.S companies going public in he period from 970 o 990, documen he underperformance of IPO s relaive o seasoned firms wih he same marke capializaion. Aggarwal and Rivoli (990) similarly find negaive afermarke performance of -3.73% in he firs year following he iniial offering for,435 IPO s in he period from 977 o 987. However, he underperformance of new issues in he afermarke has no been documened in all sudies and he inernaional evidence is varied (Loughran, e al. (994)). These inernaional variaions are due, in par, o he differences in regulaions, conracual mechanisms, and characerisics of companies going public (Firh (997)). Furher research on he long-erm sock reurn performance of IPO s and in differen marke seings seems warraned. This paper aims a () documening European IPO long-erm performance wih comparing o he U.S. IPOs; (2) invesing he sensiiviy of performance resuls o he choice of benchmark as well as he choice of mehodology; (3) idenifying, if any, he individual IPO characerisics ha explain he longerm abnormal reurn of European or U.S. IPOs. IPO characerisics include size, marke capialisaion, firs-day underpricing, indusry, capial raised, immediae pos-issuance volailiy, reained capial by he founder and year of issuance. Moreover, he sudy of he IPOs in he wo markes is very ineresing, since hey differ by he sysem of corporae governance (ousider sysem versus insider sysem). Our sample is composed of 277 companies which carried ou an IPO in Euro.NM marke beween 997 and 999. The second comparable sample is composed of 277 companies lised during he same period in NASDAQ. This pairing is carried ou by size of company a IPO dae, by indusrial secor and year of inroducion. Pairing has as a principle o neuralise he impac from he hree effecs on our resuls: he secor effec, he size effec and iming effec known as ho and cold of IPOs. Using he buy-and-hold equal-weighed mehod, our resuls for he Euro.NM sample shows ha he IPO presens a posiive long erm abnormal reurns. if we adjus his reurns by a value-weighed index, we observe a significan a long-erm underperformance of IPOs. Our resul is due o he effec of big size companies. Wih regard o our NASDAQ sample, our resuls show an underperformance of IPOs companies. The use of oher mehods o measure he long-erm performance proves he exisence of posiive abnormal reurns for Euro.NM and negaive for NASDAQ. The segmenaion of he sample by secor shows similariies for he long-erm performance of he echnology and elecommunicaion secor. For he ohers, we noe ha he performance varies from one secor o anoher. Then we ried o explain he long-erm performance for each sample by using a series of variables represening he characerisics of he company during he IPO period. This aricle is organized as follow: nex secion will review some previous sudies, mainly focusing on he sudies ha relae o long-erm performance. Secion 3 will sae he research objecives of he aricle, descripion of daa and mehodology. Secion 4 documens resuls on long-erm performance and characerisics of hree-year survival. The las secion, secion 5, draws conclusions based on he resuls in he previous secion and come ou some issues ha deserve furher sudy. 2. LONG-TERM PERFORMANCE OF IPOS The dependen variable in his analysis is he long erm performance of IPOs. There is considerable debae in he academic communiy regarding wheher underperformance exis. The purpose of his

3 research is no o explain underperformance in general or wich measuremen echniques are appropriae; raher, i aims o undersand paerns of performance in IPOs. Despie his, undersanding how measuremen affecs he findings of underperformance is useful in seing up he experimens. 2.. Empirical Evidence Three mehods have been uilized o measure he long erm performance of IPOs. Rier (99) and Loughran and Rier (995) show ha invesmen in IPOs generaes lower reurns han invesing in he marke or invesing in firms mached based on indusry and marke capializaion. Using buy-and-hold abnormal reurns (BHARs), hey examine he realized reurns of invesors who purchased each IPO in he sample period a he firs day closing price and sold afer a hree and five year o invesors in maching firms. BHARs suffer from several saisical problems. Because he reurns are aggregaed a he firm level, hey fail o accoun for he cross-correlaion in he reurns of IPOs. This is roubling, as I will show ha here is a srong cross-secional co-movemen beween IPOs ha is no explained by he Fama and French (993) hree-facor model. Also, due o he long horizon and compounding, here is an increase in variabiliy of reurns. As a resul of his, he BHARs have a righ skewed disribuion, and calculaing reliable sandard errors requires boosrapping. Also, maching firms on size alone neglecs book-omarke effecs which are predicive of fuure reurns, and IPO firms are more likely o be low book-omarke growh firms hen size mached firms which are more likely o be small due o financial disress. Because growh firms have lower expeced reurns in he FF model, his would drive a negaive bias in BHAR reurns relaive o hose firms hey are mached wih. Finally, Schulz (2003) poins ou ha, if firms are more likely o issue following IPO marke increases will cause a negaive bias as here is a higher number of issues from before a decline han afer i. Brav, e al. (2000) use a cumulaive abnormal reurn (CAR) o correc for he saisical unreliabiliy of BHARs due o compounding. Like BHARs, CARs are also aggregaed a he firm level, bu hey use he simple sum of he excess reurns from he ime following he issue. By giving equal weigh o each monh following he issue, his conrols he variabiliy of longer period reurns. Despie heir saisical properies, CARs can be an inaccurae reflecion of an invesor s realized reurn. For example, assuming marke reurns are fla, a 50% loss in one monh followed by a 00% reurn he following monh resuls in a CAR of 25%, despie he fac ha he sock is now rading exacly a is iniial price. Using he CARs and using value weighed insead of equal weighed averages grealy reduces aggregae underperformance. The CAR also does no correc for he cross-correlaion of reurns. In order o evaluae a ime series porfolio relaive o a facor model, rolling calendar ime porfolios can be used. Calendar ime average reurns (CTARs) are aggregaed by he ime period insead of he firm level. Jaffe (974) and Mandelker (974) were firs o use his mehod of analyzing sock reurns o evaluae reurns following insider rading and mergers respecively. Brav and Gompers (997) use calendar ime reurns o measure long-erm reurns following IPOs, and find ha he underperformance diminishes when his mehod is used. CTARs are useful in avoiding he saisical issues encounered wih BHARs as well as CARs. Because he reurns are aggregaed a a monhly level, he cross-secional correlaions among issuing firms are accouned for and excess volailiy due o long horizon reurns is no presen. Also, by giving an equal weighed o each monh, he Schulz (2003) pseudo marke iming bias in BHARs does no affec CTARs. Alhough CTARs are saisically preferable, hey do no have he convenien inerpreaion of a buy-and-hold sraegy reurn, and hey can also yield posiive excess reurns when socks are falling concurrenly wih he marke, even if heir fall is dramaically larger han ha of he board marke. Alhough BHARs do suffer from he saisical roubles seen above, here are wo uilizaions ha are perinen o he sudy. The firs is as an invesor-experience reurn, showing how an invesor acual wealh would have been affeced by invesing in he new issues. A second usage involves he skewed of long erm reurns which is no observable in he CTARs. Barberis and Huang (2004) poin ou ha if invesor have cumulaive prospec heory performances defined buy Tversky and Kahneman (992), hey may overweigh he small probabiliy of high success and be more willing on inves in sraegies wih an average underperformance in he ail of he disribuion Theories of Performance Miller (977) posis ha if here are consrains on shor-sellers and heerogeneous expecaions of a firm s valuaion, he sock will go o hose invesors wih he highes valuaion, and as he divergence of opinion decreases and he selling consrains are lifed, he price will fall owards he median valuaion. Duffie, e al. (2002) implemen his ino a heoreical model and drive price paerns for issues based on he consrains. As referenced earlier, Barberis and Huang (2004) argue ha because invesors may have 2

4 non-expeced uiliy preferences, lower expeced reurns may be compensaed for by a righ skewed disribuion in long erm reurns. While hese reasons ha IPOs may acually underperform he marke. Several oher explanaions may explain he underperformance seen in some works. For example, he Schulz (2003) pseudo-marke iming explanaion as well as heories in wich managers acually have he abiliy o ime he markes will predic underperformance when observaions are averaged by firms, bu no when each ime period is weighed equal. Furher, if IPOs were reflecive of a common risk facor of concern o invesors similar o size and book-o-marke facors in Fama and French (993), hen paerns of sysemaic performance would be seen, if invesor usually required a lower rae of reurn for holding new issues, his would show up as a general underperformance, when in fac he problem is ha he appropriae sock pricing model is no used in ess Cross Secion of Performance Recen IPO lieraure has urned o observing he paerns in he performance of IPOs, eiher in addiion o or insead of answering he quesion of wheher and why here is underperformance in general. Several sudies focus on issue qualiy, for example, Barberis and Huang (2004) finds underperformance only in issue wihou venure capialiss backing. Carer, e al. (998) find ha underwriers wih a beer repuaion offer issues wih lower underperformance and beer long erm performance. Neiher sudy esablishes causaion, so i is uncerain wheher a venure capialiss or higher qualiy underwrier chooses issues ha will have lower underperformance or acually conrols hese phenomena. The closes sudy o mine is ha of Krigman, e al. (999) who observe a smaller sample se ( ) and find ha soring on absolue iniial reurns, one-year reurns are increasing in iniial reurns wih he excepion of he highes iniial reurn caegory. They also find ha he higher insiuional flipping of shares predics greaer long-erm underperformance. Oher analysis focus on fricion such as in Miller (977) ha can susain a price above fundamenal valuaions as long as shoring consrains are effecive. Teoh, e al. (998) use earnings managemen proxied for by discreionary accruals, o find ha firms more aggressively managing heir earnings are able o receive a higher price for he issue hrough he IPO period, bu fall following he offering. Houge, e al. (200) use proxies for divergence of invesor opinion and finds ha, in each case, lower divergence of opinion predic less long-run underperformance. The proxies used are percenage opening spread measured by he spread a open divided by he bid/ask midpoin, he ime of he firs rade, and he flipping raio, measured as he proporion of sell-signed large block volume. 3. DATA AND METHODOLOGY This sudy focuses on iniial public offerings lised on he Euro.NM and NASDAQ. We have seleced a group of operaions in Euro.NM, and an equivalen group of operaions in he wo comparmens of NASDAQ beween 997 and 999. This sample has hus been esablished over a period of 36 monhs afer he IPO. 3.. Sources of daa and process for he selecion of he sample Firsly, we will concenrae on he 322 operaions carried ou in he differen segmens of he EuroNM (Amserdam, Brussels, Frankfor, Milan and Paris) and ha have been provided by he saisical service of he Brussels Sock Exchange. The able (panel A) successively presens he number of IPOs realized in he Euro.NM and in capial raised for he period Then, secondly, we will concenrae on he.252 operaions realized in he NASDAQ during he same period. These wo selecions have been used o make up wo paired comparable samples, one European he oher American. Thirdly, we will esablish our selecion crieria so ha our sample is no influenced by large scale IPO operaions, by specificiy or secor dominance: (i) IPOs of holding companies or banks are excluded from our sample; (ii) for each IPO in he Euro NM, we have seleced an operaion of he same size realized he same year in he NASDAQ and which belongs o he same secor of aciviy; (iii) we have eliminaed he secors which are no comparable in he wo markes; (iv) we have eliminaed he operaions which were laer eliminaed from he sock marke a few weeks afer he floaaion of he iniial quoe. Table (panel B) shows he saisics for a firs pairing, by secor of aciviy, of IPOs in he NASDAQ during he period This caegory is defined as firms wih an iniial reurn greaer han 60%. Their sample consiss of 33 issues. 3

5 Table : Saisics of our wo samples Panel A: Euro.NM sample Toal Secor N Toal (in M ) N Toal (in M ) N Toal (in M ) N Toal (in M ) Bioechnology Financial Services Indusrial & indusrial Services IT Services Media & Enerainmen Medech & Healh Care Sofware Technology Telecommunicaions Ohers wihou indexes Toal Panel B: NASDAQ sample Toal Secor N Toal (in M$) N Toal (in M$) N Toal (in M$) N Toal (in M$) Bioechnology Indusrial & indusrial Services IT Services Media & Enerainmen Medech & Healh Care Sofware Technology Telecommunicaions Toal For each of hese operaions, we had o obain he floaaion leafle for he European companies and he documens S- or he documen 424-B for he American companies. The selecion crieria cied above, were very exacing, we had o remove 7 European observaions from he financial secor. The second and hird crieria caused us o eliminae all he Media & Enerainmen secor for in Euro.NM, because 33 IPOs were realized during his period, whereas here were only 8 in NASDAQ. Finally, he necessiy of obaining he prospecus for European companies and he S- documen for he American companies obliged us o eliminae 2 supplemenary observaions. This sampling enabled us o use a group of 277 IPOs in each marke, for which, we made a sudy of he iniial low par raing, of he process of capial allocaion, of he monioring srucure and of he liquidiy of IPOs. All he observaions in our wo samples, he daa, he price, he number of shares made available o he public by he company or by is shareholders, he capial raised and he lead underwrier of he marke, have been colleced from he prospecus. The opening and closing prices, he highes, he volume deal wih, and he MTBV raio have been exraced from Daasream. The informaion abou he ownership srucure before and afer IPO is obained from he noificaion repor on ownership required by he sock marke auhoriies in each of he European and American Sock Exchanges Descripive saisics and iniial Reurns Our sample is composed of 277 companies which carried ou he ordinary IPO of shares on Euro.NM beween 997 and 999. Our benchmark sample is composed of 277 IPOs realized a he same period wih he same characerisics on NASDAQ. This pairing is carried ou by size of company a he IPO year and by indusrial secor. The objecive of pairing is o eliminae he impac of hree effecs on our resuls: he secor, size and ho and cold effecs. Table 2: Secor classificaion of he wo samples Indusry Variable Frequency Percenage Bioechnology VBSIC 5 5,42 Indusrial & indusrial services VBSIC ,30 IT services VBSIC ,80 Medech & Healh Care VBSIC 4 3 4,69 Sofware VBSIC ,02 Technology VBSIC ,47 Telecommunicaions VBSIC ,30 Euro.NM sub sample ,00 4

6 Bioechnology VBSIC 5 5,42 Indusrial & indusrial services VBSIC ,30 IT services VBSIC ,80 Medech & Healh Care VBSIC 4 3 4,69 Sofware VBSIC ,02 Technology VBSIC ,47 Telecommunicaions VBSIC ,30 NASDAQ sub sample ,00 Global sample 554 Table 2 presens he secor disribuion of our wo samples during our empirical sudy. This able presens he secor classificaion according o he Euro.NM s auhoriies. For our NASDAQ sample, he classificaion is based on SIC code (Sandard Indusrial Classificaion) which is used by he auhoriies of his marke o accep or refuse he enry of a company in he calculaion of he indices. This classificaion enables us o use binary variables accordance wih he mehodology of Lee, e al. (993). Table 3: Descripive saisics of our Euro.NM IPOs and he paired IPOs on NASDAQ beween 997 and 999 Panel A. IPO characerisics of Euro.NM sample (N=277) Mean Median Sandard deviaion Min Max Skew IPO Volume ,8 New Shares (% of IPO) 78,63 80,90 2,67 0,00 00,00 -,6 Old shares (% of IPO) 2,45 9,42 2,64 0,00 00,00,6 Green-Shoe 23,778 47, , ,87 IPO price ( ) 25,73 2,00 35,66 0,76 559,87 2,37 IPO size (en millions d ) 34,89 20,90 44,39 2,96 447,90 4,50 Marke value a IPO (in M ) 23,39 72, ,2 899,50 2,68 Panel B. IPO characerisics of NASDAQ sample (N=277) Mean Median Sandard deviaion Min Max Skew IPO Volume ,93 New Shares (% of IPO) 94,9 00,00 2,27 26,32 00,00-2,73 Old shares (% of IPO) 5,22 0,00 2,40 0,00 76,68 2,67 Green-Shoe ,72 IPO price ($) 0,9 9,50 4,43 3,50 30,25, IPO size (M ) 36,26 29,7 3,27 3,50 87,00,85 Marke value a IPO (in M ) 57,79 99,60 76,72 9,7.053,3 7,54 Table 3 shows he characerisics of he IPOs for he wo samples. Our resuls show ha he IPO volume of he NASDAQ s companies is more imporan han hose of he Euro.NM. Moreover, in order o ensure a high level afer marke liquidiy, hese companies fix a low IPO price. The average IPO size for Euro.NM is 35 millions euros and 36 millions dollars. Finally, able 3 shows ha he old shareholders of Euro.NM companies offer an average of 2.45% of he IPO. On he oher hand, hose of our NASDAQ sample ake par only of 5.22% in he operaion. These companies prefer o increase he capial, conrary o he Euro.NM companies, where heir shareholders end o privilege he immediae liquidiy. Table 4 presens he saisics on IPO underpricing for he 544 observaions. The average marke adjused reurns (MAR) observed he firs rading day on our Euro.NM sample is higher han ha observed on NASDAQ. Wih an average adjused reurn of 30,84%, he Ialian segmen of Euro.NM is he highes. On he oher hand, he Belgian segmen presens he lower average reurn (4,0%) wih a endency o become negaive afer hree weeks following he IPO dae. However hese wo segmens represen only 6% of our Euro.NM sample. The German segmen represens he second higher average (5.4%) afer he Ialian marke, in spie of he fac ha i represens 60 % of our Euro.NM sample. The average marke adjused reurn observed on he segmen of Paris is only 25.83%. I is even lower han he average observed on he wo comparmens of NASDAQ which is 33.68% for NASDAQ NNM and 9.58 % for NASDAQ SCM. The averages adjused or non adjused reurns are posiive for he boh samples. For he European IPO sample, he average marke adjused reurns is 43.99% a he significan level of 0.0. On he oher hand, for our NASDAQ sample, he average is 30.57% a he significan level of 0.0. These resuls prove ha i is more ineresing for he shareholders o carry on an IPO on he NASDAQ marke han on Euro.NM. Our resuls corroborae he resuls of Dewener and Malaesa (997). 5

7 Table 4: Abnormal reurns observed on he Euro.NM and NASDAQ beween 997 and 999 This able presens he average and he median of he abnormal adjused and non-adjused reurns IPOs for he whole of he period on he various segmens of Euro.NM (Amserdam, Brussels, Frankfur, Milan and Paris) and he paired IPOs carried ou on he NASDAQ for he same period. The oupus are measured over various periods: s, 7h, 2s, 30h, 60h and 90h day of he negoiaions. The non adjused reurns are compued according o equaion 2 and marke adjused reurns according o he equaion 3. α, β and γ indicae respecively he significan levels o he hreshold of 0%, 5% and % of he Suden es-saisics. The es is carried ou o es if he average of reurns is differen from zero. I is esimaed by he raio: mean/sandard deviaion; where, he sandard error represens he sandard deviaion divided by he square roo of he number of observaions. Mean; Median (-saisic; simple size) Non-adjused reurns (%) Marke adjused reurns (%) Marke s day 7 h day 2 s day 30 ème day 60 ème day 90 h day s day 7 h day 2 s day 30 h day 60 h day 90 h day All Euro.NM sample 43,98 ; 9,35 (9,49 γ ; 277) 48,20 ; 20,40 (9,73 γ ; 277) 56,04 ; 24,94 (0,85 γ ; 277) 6,27 ; 25,52 (0,82 γ ; 277) 93,7 ; 29,63 (8,84 γ ; 277) 3,86 ;43,08 (8,07 γ ; 277) 43,99 ; 9,32 (9,50 γ ; 277) 45,78 ; 8,02 (9,29 γ ; 277) 49,67 ; 7,42 (9,95 γ ; 277) 5,78 ; 20,66 (9,64 γ ; 277) 72,42 ; 2,34 (7,70 γ ; 277) 96,64 ; 27,40 (6,70 γ ; 277) Germany 5,30 ; 24,02 (8,48 γ ; 67) 52,70 ; 28,93 (8,82 γ ; 67) 60,06 ; 32,52 (9,95 γ ; 67) 66,32 ; 39,09 (9,79 γ ; 67) 0,52 ; 52,27 (8,34 γ ; 67) 36,44 ; 64,2 (8,56 γ ; 67) 5,4 ; 25,84 (8,48 γ ; 67) 50,82 ; 27,03 (8,57 γ ; 67) 54,97 ; 30,38 (9,33 γ ; 67) 58,49 ; 35,3 (8,99 γ ; 67) 80,44 ; 35,59 (7,57 γ ; 67) 99,57 ; 42,60 (7,34 γ ; 67) Belgium 3,7 ;,6 (0,75 ; 0),95 ; -2,06 (0,29 ; 0) -2,78 ; -,89 (-0,58 ; 0) -6,9 ; -5,85 (-0,95 ; 0) -3,90 ; -8,38 (-0,40 ; 0) -8,92 ; -4,35 (-0,74 ; 0) 4,0 ; 4,94 (0,92 ; 0) 0,07 ; -,55 (0,0 ; 0) -6,5 ; -6,3 (-,59 ; 0) -9,2 ; -9,9 (-,48 ; 0) -8,56 ; -0,58 (-0,99 ; 0) -6,9 ; -28,9 (-,57 ; 0) France 25,95 ; 0,3 (5,90 γ ; 80) 29,86 ; 8,76 (4,7 γ ; 80) 38,62 ; 2,07 (4,7 γ ; 80) 42,5 ; 9,69 (4,38 γ ; 80) 64,73 ; 0,74 (3,52 γ ; 80) 98,87 ; 2,09 (3,40 γ ; 80) 25,83 ; 9,47 (5,89 γ ; 80) 26,87 ; 6,80 (4,30 γ ; 80) 3,7 ; 9,37 (3,5 γ ; 80) 30,9 ; 6,4 (3,45 γ ; 80) 45,4 ; 5,86 (2,78 γ ; 80) 68,98 ; 9,63 (2,78 γ ; 80) Ialy 32,06 ; 2,62 (,39 ; 7) 43,82 ; 94,35 (2,45 α ; 7) 28,93 ; 247,7 (4,27 γ ; 7) 25,53 ; 245,9 (4,3 γ ; 7) 422,76 ; 235, (2,43 α ; 7) 739,30 ; 264,9 (2,5 α ; 7) 30,84 ; 20,42 (,39 ; 7) 34,5 ; 8,53 (2,25 α ; 7) 9,76 ; 27,8 (3,82 γ ; 7) 58,8 ; 7,23 (3,58 β ; 7) 335 ; 422,76 (2,00 α ; 7) 593,27 ; 76,8 (,75 ; 7) Neherlands 44,76 ; 29,87 (4,5 γ ; 3) 34,09 ; 45,29 (,98 α ; 3) 69,08 ; 3,22 (2,44 β ; 3) 64,00 ; 27,08 (2,58 β ; 3) 69,57 ; 28,09 (2,42 β ; 3) 57,27 ; 44,4 (2,26 β ; 3) 44,32 ; 28,72 (4,3 γ ; 3) 27,83 ; 37,4 (,87 ; 3) 6,76 ; 25,83 (2,4 α ; 3) 53,22 ; 7,09 (2,08 α ; 3) 57,68 ; 22,55 (,95 α ; 3) 48,66 ; 30,29 (,78 α ; 3) All NASDAQ marke 30,69 ; 0,00 (8,3 γ ; 277) 28,65 ; 8,69 (8,0 γ ; 277) 38,26 ;, (7,60 γ ; 277) 33,92 ; 8,80 (6,64 γ ; 277) 38,54 ; 7,60 (5,92 γ ; 277) 5,53 ; 8,44 (6,00 γ ; 277) 30,57 ; 9,86 (8,30 γ ; 277) 27,97 ; 8,45 (8,02 γ ; 277) 36,2 ;,85 (7,40 γ ; 277) 30,36 ; 6,49 (6,7 γ ; 277) 30,45 ; 2,8 (4,84 γ ; 277) 38,36 ; -2,50 (4,60 γ ; 277) NASDAQ NNM 33,85 ; 0,7 (7,53 γ ; 26) 32,28 ; 0,2 (7,49 γ ; 26) 45,75 ; 2,50 (7,32 γ ; 26) 4,22 ; 0,65 (6,49 γ ; 26) 50,9 ; 7,98 (6,2 γ ; 26) 66,65 ; 22,92 (6,32g ; 26) 33,68 ; 0,29 (7,5 γ ; 26) 3,5 ; 0, (7,4g ; 26) 43,46 ; 5,04 (7,4 γ ; 26) 37,3 ; 8,62 (6,08 γ ; 26) 4,42 ; 0,69 (5,30 γ ; 26) 52,76 ; 0,23 (4,5 γ ; 26) NASDAQ SCM 9,53 ; 9,33 (3,84 γ ; 6) 5,8 ; 7,33 (3,36 γ ; 6) 0,53 ; 4,08 (2,67 γ ; 6) 8,05 ; 5,00 (,92 α ; 6) -2,7 ; -7,8 (-0,60 ; 6) -2,00 ; -2,40 (-0,25 ; 6) 9,58 ; 9,36 (3,87 γ ; 6) 5,40 ; 6,32 (3,35 γ ; 6) 0,53 ; 4,08 (2,74 γ ; 6) 5,75 ; -,25 (,52 ; 6) -8,42 ; -9,55 (-,90 ; 6) -2,64 ; -9,86 (-,60 ; 6) 6

8 3.3. Selecion of he conrol porfolios We measure he long-run performance of our sample IPOs beween 997 and 999, using coninuously rebalanced and purged conrol porfolios (size and/or marke-o-marke raios). We consiue hree ses of benchmark porfolios, in he same way used by Barber and Lyon (997). The firs se of conrol porfolios is consiued by five porfolios reconsiued every year in July. For June of each year, we classify all he Euro.NM s (he NASDAQ s) companies according o heir size, measured by he marke capializaion. Then afer he Euro.NM s (he NASDAQ s) companies are classified in heir quinile of suiable size, based on he marke value of he share for June. We compue he monhly reurn for each porfolio by using he equal weighed average of all shares belongs o he same quinile of size. In June each year, we classify he porfolios and he companies are auhorized o change once per year he size quinile. The size-benchmark reurn is equivalen o a sraegy of invesmen in a size weighed porfolio wih a monhly rebalancing. The second whole of reference porfolios is composed of five porfolios reconsiued according o he level of MTBV raio (July of each year). December of he year T -, we classify all he companies of Euro.NM (of NASDAQ) in various populaions according o heir level of MTBV raio. Then afer, we consiue quiniles based on he MTBV raios for all he companies of Euro.NM (of NASDAQ). Finally, he Euro.NM (NASDAQ) companies are placed in heir suiable MTBV quinile while being based on he MTBV value of he year T. The reurns on five MTBV porfolios are calculaed in a way similar o he five size-porfolios. Our hird se of reference porfolios is composed of 25 Size/MTBV porfolios which are reconsiued in July of each year. These porfolios are made up in wo seps. In he firs sep, June of he year T, we classify all he companies of Euro.NM in our sample on he basis of sock exchange capializaion of he share. Then, we consiue quiniles by basing us on hese classificaions of all he Euro.NM companies. In he second sep, wihin each Size quinile, he companies are classified in quiniles according o values of MTBV raios during he year T. The companies of Euro.NM are placed in heir suiable Size/MTBV porfolio based on heir size during June of he year T and on he value of heir MTBV raio for he year T. The reurns of he 25 porfolios are compued in a similar way o ha of he five Size and MTBC porfolios. Finally, in addiion o he hree ses of reference porfolios, we ake he Euro.NM All-shares equal weighed index ( NASDAQ Composi for he NASDAQ sample). We also compue a value weighed index porfolio. The IPOs are assigned o each porfolio and heir reurn is compared wih ha of he porfolio o deermine he abnormal reurn. The classificaion of he companies in Size and MTBV porfolios he monh which follows he IPO is presened in able 5. Table 5: Classificaion of companies in porfolios according o heir sizes and Markeo-Book raios Panel A : Disribuion of he Euro.NM sample Quiniles «Size» Quiniles Marke-o-Book Value Low Quinile 2 Quinile 3 Quinile 4 High Toal Small Big Toal Panel B : Disribuion he NASDAQ sample Quiniles «Size» Quiniles Marke-o-Book Value Low Quinile 2 Quinile 3 Quinile 4 High Toal Small Big Toal The able shows ha % of he IPOs on NASDAQ (Panel B) have a high MTBV. On he oher hand, only 52.7 % of he Euro.NM IPOs (Panel A) are companies wih high MTBV raio. Moreover, here is no company lised on he Euro.NM wich characerized by a high MTBV raio and wih a small size. 7

9 3.4. Iniial Public Offering Performance Measuremen We have calculaed he abnormal reurns for IPOs in he periods of 6, 2, 8, 24, 30 and 36 monhs. The choice of hese differen ime scales enabled us o examine he long-erm behaviour of several caegories of invesor. Numerous recen sudies have analysed long-erm abnormal reurns by using differen mehods. More recenly, Barber and Lyon (997), Kohari and Warner (997), Lyon, e al. (999), Fama (998), Loughran and Rier (2000), Brav, e al. (2000) and Michell and Safford (2000), have all demonsraed ha he mehod for measuring abnormal reurns influences boh he size and he srengh of he saisical es. Given ha each of hese measuring mehods used in he lieraure has, up o now, shown is limiaions, we will use all he mehods for our research. Thus, we will be able o examine he long-erm performance of IPOs by referring o a variey of models. We will rely on he papers of de Barber and Lyon (997), Kohari and Warner (997), Fama (998) and Lyon, e al. (999), and we will use four measures o evaluae he long-erm performance of iniial public offerings. To calculae he afermarke long-erm performance, Loughran and Rier (2000) exclude from heir calculaions he firs day reurns. However, we consider ha he abnormal behaviour of IPOs is correlaed o he phenomenon of under-pricing. In order o disinguish he valuaion error made by he invesors during he firs marke day o ha commied by he lead underwrier, we sugges ha afermarke performance should also be measured by using he IPO price. On he one hand, his procedure will enable us o observe he afermarke performance of he offers ofen acquired by insiuional invesors who have he privilege of buying a he subscripion price. On he oher hand, i will enable us o examine he afermarke performance of hose acquired by individual invesors a he marke price Cumulaive Average Adjused Reurns (CAR) The adjused abnormal reurn, AR i,, for he company i over a period of calendar monhs following he firs rading monh is calculaed in he following manner: AR i, = R i, E(R i,benchmark ) () Where R i, is he reurn for firm i in even monh and E(R i,benchmark ) is he reurn on he benchmark during he corresponding ime period. The average benchmark-adjused reurn on a porfolio of n socks for even monh is he equally-weighed arihmeic average of he benchmark-adjused reurns: AR = n n i= AR i, The cumulaive benchmark-adjused reurn for he afermarke performance from even monh q o even monh s, CAR i,q (ha implicily supposes he monhly porfolio rebalancing) is he summaion of he average benchmark-adjused reurns: CAR, AR (3) = S q s = q The saisical es carried ou on he cumulaed abnormal reurns is obained by using he following formula: CAR CAR i, = (4), σ ( CARi, ) / n Where σ(car i, ) is he cross-secional sample sandard deviaions of abnormal reurns for he sample of n firms and n is he number of IPOs on monh. Following Barber and Lyon (997), we prefer he use of cross-secional sandard errors because requiring pre-even reurn daa, from which a imeseries sandard errors can be esimaed, inensifies he new lising bias. More specifically, he saisical es for he CAR, is: CARi n CAR = (5), [ var + 2 ( ) cov Where var is he average of he cross-secional variaions over 36 monhs of he AR i, and Cov is he firs order auo-covariance of he AR series. (2) 8

10 Buy-and-Hold Abnormal Reurns (BHAR) The second measure we use is based on he calculaion of he T holding period 2 reurn as an alernaive o he use of he cumulaive benchmark-adjused reurns (no porfolio rebalancing is assumed in hese calculaions), defined as: = T, T + i, ) = R i ( r (6) This measure makes i possible o calculae he oal reurns procured by a sraegy called Buyand-Hold in which a share acquired a he closing price on he firs rading day is reained up o monh T afer he IPO dae. The average Buy and Hold reurns (no rebalancing is assumed in his calculaion) for all he companies in each of our wo samples, during he monh T, is simply equal o he average of he reurns of each firm in he same period: R = n T R i, T n i= Where n is he number of companies in he sample. The abnormal buy and hold reurns adjused from he normal performance of he reurns rae E(R benchmark, ) over he same period is defined by: T T BHARi, T = + ri, ) ( + E( Rbenchmark, ) = = ( (8) The average of adjused abnormal reurns for he period is defined by: BHAR = n i= x BHAR (7) i, i, (9) The weigh x i, is /n when abnormal reurns are equally-weighed and n MV i / MV i when abnormal reurns are value weighed, MV is he marke value and n is he number of companies during he period. The null hypohesis H 0 saes ha he BHAR for all he companies in each of our wo samples for he monh T is equal o zero: H 0 : BHAR T = 0 To es he null hypohesis, we prefer o use he saisical es adjused from he skewness recommended by Neyman and Pearson (928) and recenly used by Lyon, e al. (999). The es is defined by: ) 2 ) = n ( S + γs + γ ) (0) 3 6n Where: Moyenne ( BHAR) S = ; = 6, 2, 8, 24, 30 and 36 monhs ; ) γ is an esimaor of he coefficien σ ( BHAR) of he skewness : ˆ γ = n i= ( BHAR i BHAR ) nσ ( BHAR ) The calendar-ime porfolio mehods Loughran and Rier (995) and Brav and Gompers (997) use Fama-French s hree- facor model o measure he reurns in he Calendar-Time Porfolios of IPOs. Jaffe (974) and Mandelker (974) use several of hese mehod ypes. As well as he CARs and he BHARs, mehod, we will consider as a hird alernaive, wo ypes of mehods among he Calendar-Time Porfolio : he firs, based on he use of he hree-facor models developed by Fama and French (993) and he second based on he monhly average of he Calendar-Time Abnormal Reurns. Fama (998) and Lyon, e al. (999) confirm ha he Calendar-Time Porfolio mehods offer wo advanages. The firs is ha i eliminaes he problem of cross-secional dependence beween he reurns of he companies in he sample. The second is ha hey make he es saisics more robus on he i= 2 Roll (983, p. 377) poin ou ha buy-and- hold mehod ( ) gives an unbiased esimae of he holding period reurn on a realisic porfolio. Barber and Lyon (997) also prefer o use his mehodology. They confirm ha his is he bes mehod for sudying he long erm behaviour of he invesor. These auhors criicise he use of he CAR mehod for a long-period. In fac, in heir opinion, he mehod is robus for measuring shor-erm reurns, bu i is a biased esimaor in he conex of long-erm abnormal reurns. 9

11 samples. In he nex wo secions, we will presen he mehodological procedure ha we have followed o apply hese wo mehods. A. Fama-French s hree facor model Le us suppose ha he even period is of hree years. For each calendar monh, we calculae he reurns on porfolio made up of companies which have an IPO on he sock marke in he las hree calendar years. The Calendar-Time reurns in his porfolio are used o esimae he following regression: R p R f = a i + β i [R m R f ] + s i SMB + h i HML + e i () Where R p represens he porfolio of sock marke lisings and includes all he IPOs beween 997 and 999; (R m R f ) represens he excess of marke reurn in relaion o ha of he free risk asse. The firs erm, R m represens he average reurns of he securiies making up he index Euro.NM weighed by he sock marke capialisaion of each securiy. We also use an equally-weighed average reurn. The second erm R f represens he free risk asse reurns, he EUROR hree monh rae. 3 SMB (Small Minus Big) is he difference each monh beween he average reurns in he hree small porfolios and he average reurns in he hree large porfolios. 4 SMB = /3(Small Value + Small Neural + Big Value) /3(Big Value + Big Neural + Big Growh). HML (High Minus Low) is he difference each monh beween he average reurns of he porfolio wih a high MTBV raio and he average reurns of he porfolio wih a low MTBV raio, HML = /2(Small Value + Big Value) /2(Small Growh + Big Growh). α i, β i, s i and h i represen he parameers for esimaing he regression equaion. The esimaion of he consan α of he regression enables us o es he null hypohesis according o which, he monhly average of he reurn surplus in he Calendar-Time Porfolio is equal o zero. The inerceps in hese regressions can be inerpreed in a similar way o Jensen s alpha in he conex of he work on he CAPM. Given ha he number of securiies which consiue he Calendar-Time Porfolio vary from one monh o anoher, he disribuion of he error erm (e ) may be Heeroscedasic. To overcome his problem, according o Boehme and Sorescu (2002) proposiion, we esimae he equaion using a Weighed Leas Square (WLS). The weighing facor is based on he square roo of he number of securiies making up he porfolio in each calendar monh. B. The Calendar-Time Abnormal Reurns (CTAR) Le us suppose ha he even period is hree years. For each calendar monh, we calculae he abnormal reurns (AR i ) for each securiy i by using he reference porfolio reurns (R p ) over he same period: AR i, = R i, R p (2) For each calendar monh, we have calculaed he mean reurns (MAR T ) across firms in he porfolio over he las 6, 2, 8, 24, 30 and 36 monhs, ha is o say, we had o recreae he porfolio each monh: MAR T n = i= x AR i, i, (3) Where n is he number of companies in he porfolio during he monhs,, x i is he weigh of he abnormal reurns, equal o /n if hey are equal-weighed and equal o n MV i / MV i if hey are valueweighed. The number of Calendar-Time porfolio varies from one monh o anoher. If during a paricular monh, he porfolio does no conain any firms, we did no use ha monh. The monhly MAR is sandardized by using he porfolio sandard deviaion porfolio as an esimaor. Michell and Safford (2000) evoke wo reasons for such a procedure. Firsly, i makes i possible o conrol he heeroskedasiciy. Then, i makes i possible o place more imporance on he periods characerised by grea even aciviy in comparison wih period of low aciviy. 5 Then, we calculae grand mean monhly abnormal reurns (MMAR) using he sandardised MAR T : MMAR = T T = MAR( S an dardisé) Where T are he oal number of calendar monhs. In order o es he null hypohesis of zero mean monhly abnormal reurns, a saisic is calculaed using he ime-series sandard deviaion of he mean monhly sandardized abnormal reurns: i= (4) 3 For he NASDAQ sample, we use he reurns in he NASDAQ composie (R m ) and he rae of he hree monh Treasury bills (R f ). 4 For a more deailed descripion of he creaing of hese porfolios see Fama and French (993). 5 Everyhing being equal elsewhere, he porfolio variance increases according o he size of he porfolio. 0

12 4. EMPIRICAL RESULTS MMAR ( MMAR) = (5) σ [ MAR ( normalisé ) T ] / T 4.. The Afermarke Performance of Iniial Public Offerings 4... Resuls by using he cumulaed abnormal reurns (CAR) Table 6 presens he average of non-adjused reurns (R ) and he average of cumulaed abnormal reurns (CAR, ) for he 36 monhs ha follows he IPO dae. The daa in Panel A indicaes he resuls for 277 IPOs realized on he Euro.NM during he period Panel B shows he resuls for he 277 equivalen IPOs made in he NASDAQ during he same period. Table 6: IPO Abnormal reurns according o he CAR mehod Unadjused reurns Monhs Number R CR, 5 porfolios "MTBV" 5 porfolios "Size" porfolios "Size-MTBV" Euro.NM Index (EW) Euro.NM Index (VW) afer IPO of IPO % - saisic % - saisic % - saisic % - saisic % - saisic % - saisic % - saisic 277 7,60 3,56 7,60 4,33 4,90 3,20 4,8 3,9 5,02 3,2 3,80 2,53-2,4 -, ,76 4,6 5,35 6,2 0,23 4,72 9,95 4,65 0,69 4,68 6,7 3,5-5,56-2, ,80 4,23 24,5 7,83 6,48 6,20 5,78 6,02 6,87 6,03 0,58 4,05-7,4-2, ,87 4,80 32,02 8,97 9,83 6,46 8,33 6,05 20, 6,22,96 3,97-2,2-3, ,83 2,98 37,86 9,47 23,76 6,92 2,02 6,2 23,93 6,62 3,65 4,05-6,78-4, ,32 3,45 46,8 0,54 29,69 7,90 26,8 7,06 29,33 7,4 7,25 4,67-9,6-5, ,23 3, 5,42 0,86 29,89 7,36 25,08 6,26 27,5 6,35 5,22 3,82-29,96-7, ,72 4,03 58,4,48 3,85 7,34 25,6 5,98 28,70 6,28 4,59 3,42-37,32-8, ,9 2,97 63,33,79 32,75 7, 24,66 5,43 28,73 5,93 2,84 2,84-45,75-9, ,36 4,5 7,70 2,66 38,48 7,93 29,28 6, 32,98 6,45 6,8 3,39-47,4-9,58 277,76 0,87 73,46 2,36 39,85 7,83 30,05 5,98 34,67 6,47 5,77 3,5-5,09-9, ,92 -,9 7,53,52 39,99 7,52 29,43 5,6 33,53 5,99 4,20 2,72-57,29-0, ,80-2,47 67,74 0,48 39,04 7,05 28,02 5,3 32,72 5,62,22 2,06-64,52 -, ,48 -,70 65,26 9,73 38,42 6,69 26,83 4,73 32,38 5,36 8,83,57-69,93-2, ,2-0,3 65,05 9,37 38,76 6,52 26,33 4,49 32,84 5,25 7,44,27-73,99-2, ,69-0,89 63,36 8,83 40,66 6,62 27,75 4,58 34,88 5,40 7,34,22-77,59-2, ,22-3,4 58,3 7,86 39,20 6,9 25,67 4, 33,09 4,97 4,56 0,73-85,20-3, ,9-2,94 53,22 7,00 36,02 5,53 2,72 3,38 29,79 4,35-0,53-0,08-95,7-4, ,94-0,57 52,29 6,66 34,74 5,7 9,44 2,93 28,32 4,00-3,69-0,56-02,75-5, ,3-0,76 50,97 6,33 34,59 5,02 8,80 2,77 28,33 3,9-6,3-0,9-09,0-5, ,25 -,66 47,73 5,79 35,80 5,07 9,74 2,83 29,74 4,00-6,53-0,94-2,33-5, ,58-4,87 4,4 4,87 34,33 4,75 8,02 2,53 28,40 3,73-9,28 -,3-7,97-6, ,90 -,66 38,24 4,43 36,24 4,90 9,77 2,7 30,50 3,92-8,53 -,8-20, -6, ,05 -,27 36,9 4,0 37,53 4,96 20,83 2,79 30,93 3,89-7,95 -,07-23,5-6, ,7-0,87 34,48 3,8 39,80 5,4 22,97 3,00 33,09 4,06-6,22-0,82-25,44-5, ,03-2,22 3,45 3,40 39,37 4,97 22,2 2,83 32,33 3,88-7,6-0,92-29,2-6, ,66-5,0 23,79 2,52 37,52 4,64 9,72 2,47 29,94 3,52-0,57 -,33-34,49-6, ,24 -,95 20,55 2,0 40,00 4,77 2,88 2,64 32,88 3,72-9,22 -,2-35,50-5, ,57-4,22 3,98,33 36,83 4,09 8,88 2,3 29,74 3,4-3, -,48-43,25-5, ,9-2,25 0,07 0,9 36,85 3,90 8,85 2,02 30,20 3,03-4,24 -,53-48,07-5, ,66-0,32 9,4 0,83 38,8 3,92 20,5 2,0 3,00 3,03-3,45 -,4-49,39-5, ,95-0,36 8,46 0,72 40,56 4,06 22,23 2,25 33,97 3,23 -,62 -,8-50,33-4, ,92 -,32 5,54 0,45 42,55 4,0 24,20 2,3 35,98 3,22-0,46 -,00-5,43-4, ,36 -,03 3,8 0,24 43,46 3,78 25,45 2,24 37,35 3,09-9,84-0,87-53,38-3, ,24-2,33 -,07-0,07 42,59 3,46 24,45 2,0 36,40 2,8 -,4-0,94-57,47-2, ,38 -,95-5,45-0,36 43,28 3,34 24,94,95 36,33 2,66 -,04-0,87-60,55-2,23 Unadjused reurns Panel A. resuls for he Euro.NM sample Panel B. Resuls for he NASDAQ sample Cumulaive abnormal reurns (CAR,) Cumulaive abnormal reurns (CAR,) Monhs Number R CR, 5 porfolios "MTBV" 5 porfolios "Size" Porfolios "Size-MTBV" Nasdaq Index (EW) Nasdaq Index (VW) afer IPO of IPO % - saisic % - saisic % - saisic % - saisic % - saisic % - saisic % - saisic 277,00 0,60,00 0,45 -,22-0,60 -,06-0,5 -,4-0,70-0,79-0,39 -,39-0, ,85-0,50 0,5 0,05-4,77 -,67-3,96 -,35-4,43 -,55-4,02 -,39-5,36 -, ,68,90 3,83 0,99-4,62 -,32-2,78-0,77-3,85 -,0-3,7-0,89-5,54 -, ,46 2,4 9,29 2,08-2,98-0,73-0,56-0,4-2,3-0,57 -,26-0,3-4,29 -, ,4 2,75 4,43 2,89-2,3-0,47 0,72 0,5 -,25-0,28 0,0 0,02-4,28-0, ,98,75 7,4 3,9-2,4-0,43 0,86 0,7 -,43-0,29 0,39 0,08-4,58-0, ,98,60 20,38 3,44-3,77-0,70-0,30-0,06-3,4-0,59-0,42-0,08-6,36 -, ,07,8 22,45 3,52-4,68-0,8-0,93-0,6-3,65-0,63 -,3-0,9-7,68 -, ,0,5 25,56 3,77-4,55-0,74-0,42-0,07-3,28-0,53-0,5-0,02-7,7 -, ,26-3,77 9,29 2,68-0,58 -,62-6,24-0,93-9,6 -,4-5,96-0,90-3,47-2, ,33-2,82 4,96,98-6,44-2,40 -,72 -,67-4,82-2,7 -,39 -,64-20,27-2, ,73 -,47 2,23,55-8,76-2,62-4,57 -,99-7, -2,40-4,6 -,96-22,55-3, ,7-2,6 7,52 0,9-23,73-3,6-9,92-2,60-22,59-3,02-9,27-2,54-28,63-3, ,38,5 0,90,27-22,87-2,93-8,84-2,37-22,2-2,86-8,33-2,33-27,90-3, ,3-0,06 0,77,2-23,73-2,93-9,95-2,4-23,07-2,86-9,37-2,37-29,08-3, ,7-0,92 8,60 0,93-25,39-3,03-22,08-2,58-24,94-2,99-2,6-2,55-3,03-3, ,83 -,46 5,77 0,60-28,64-3,30-25,2-2,84-28,32-3,28-24,88-2,84-35,62-4, ,82,79 0,60,06-25,80-2,85-2,80-2,36-25,87-2,87-22,50-2,46-3,96-3, ,6,09 3,2,29-25,22-2,7-2,74-2,29-26,6-2,87-22,0-2,35-3,30-3, ,53,35 6,73,58-23,6-2,46-20,32-2,07-25,54-2,67-2,52-2,22-29,04-3, ,72 0,75 8,46,70-25,07-2,54-2,72-2,6-27,59-2,8-24,80-2,49-29,73-3, ,05 0,92 2,5,93-22,89-2,26-20, -,95-25,93-2,57-24,37-2,38-26,69-2, ,57,64 26,09 2,28-20,75-2,00-8,22 -,72-24,85-2,40-24,40-2,32-24,48-2, ,56 0,25 26,65 2,28-22,49-2,2-20,35 -,88-27,36-2,59-27,22-2,54-25,5-2, ,48 2,0 32,3 2,66-20,99 -,92-9,3 -,73-25,8-2,37-27,46-2,48-23,05-2, ,93,35 35,06 2,83-20,65 -,83-2,27 -,85-26,3-2,34-28,93-2,54-2,94 -, ,39,30 38,45 3,02-20,9 -,75-2,65 -,83-26,62-2,3-3,35-2,69-20,70 -, ,29 -,4 35,6 2,66-25,5-2,0-28,88-2,36-32,30-2,70-38,78-3,20-23,06 -, ,75 0,60 36,9 2,70-26,97-2,7-3,26-2,47-34,96-2,83-42,57-3,39-24,35 -, ,04 2,5 42,96 3,02-24,44 -,89-28,77-2,8-32,37-2,5-4,00-3,4-7,97 -, ,0,55 46,97 3,22-23,76 -,79-29,28-2,6-33,36-2,53-4,48-3,0-6,30 -, ,88 0,6 48,85 3,24-23,75 -,73-29,32-2,0-34,24-2,5-4,83-3,02-3,09-0, ,29,26 52,4 3,32-24,72 -,73-30,72-2, -37,09-2,6-44,5-3,09-3,3-0, ,54,6 56,68 3,46-20,99 -,4-29,07 -,92-34,55-2,33-4,29-2,75-6,9-0, ,69 2,23 63,37 3,72-8,86 -,22-26,45 -,67-3,98-2,07-38,9-2,49-2,94-0, ,46 0,2 63,82 3,62-8,49 -,6-28,35 -,74-32,38-2,03-39,33-2,43 0,69 0,04 By using he reference porfolio for he adjusmen of he reurns, i appears, a firs sigh, ha he companies ha floa shares in he sock marke in he Euro.NM do no show a decline in heir performance. This observaion seems o go agains our resuls for equivalen NASDAQ companies as well as he resuls of Rier (99) e Loughran and Rier (995). In fac, of he 36 non-adjused average reurns calculaed for he companies in he Euro.NM, he firs observaions showed posiive signs. Apar from hese high reurns, he decline in he performance of companies of he size and / or he same MTBV raio has repercussions on he abnormal reurns and gives a posiive cumulaive average reurn over he 36 monhs of he sudy. The use of he Euro.NM All Shares EW and VW indexes qualifies hese resuls and does no make i possible o reach a conclusion concerning he coninually high sock marke performances of Euro.NM companies. As for our NASDAQ sample, he resuls show, on he whole, a durable decline in he performance of companies wih a sock marke floaaion during his

13 period. By adjusing he reurns of his sample by 25 Size MTBV porfolios, we find an average of cumulaive abnormal reurns of % over 36 monhs. 6 These resuls are calculaed from he closing price observed on he firs rading day. We will now ake ino consideraion he subscripion price in he floaaion offer. In figure, we graphically represen he cumulaive abnormal reurns for our wo samples. We have calculaed using several reference porfolios. The iniial non-adjused average reurn is 56.04% on he Euro.NM, followed by a monhly average reurn ha varies beween and 7.66%. The average cumulaed reurns reach a maximum of 2.9 % he elevenh monh, hen decrease. This decline can parly be aribued o he speculaive bubble which has affeced he echnological values during his period. As for he NASDAQ companies, we observe a lower iniial reurn han ha observed in he Euro.NM. We find an iniial reurn of 38.26%, followed by a monhly average reurn ha varies beween and 6.26%. The cumulaive abnormal reurns also reach a high level of 0.08% in he 36h monh and seem o coninue afer ha. However, his increase is much lower and slower han ha observed on our Euro.NM sample. This difference can be explained by he fac ha he speculaive bubble is much larger in he NASDAQ han in he Euro.NM. Figure : Cumulaive abnormal reurns from subscripion price 20,00 Cumulaive Abnormal Reurns (CAR en %) 00,00 80,00 60,00 40,00 20,00 0, Cumulaive Abnormal Reurns (% CAR) 00,00 80,00 60,00 40,00 20,00 0, ,00-40,00-60,00-80,00-00,00-20,00 Monhs afer lising on Nasdaq -20,00 Monhs afer lising on Euro.NM Unadjused Abnormal Reurns Adjused by porfolio: MTBV Adjused by porfolio: Size Unadjused Abnormal Reurns Adjused by porfolio: MTBV Adjused by porfolio: Size Adjused by porfolio: Size-MTBV Adjused by Nasdaq Composie Index (EW) Adjused by Nasdaq Composie Index (VW) Adjused by porfolio: Size-MTBV Adjused by Euro.NM Index (EW) Adjused by Euro.NM (VW) Index Apar from he hree reference porfolios, figure, reraces he evoluion of he cumulaed abnormal reurns adjused by he marke reurn. Each monh, we subrac he reurn observed in he marke from he reurn of each securiy, by using wo indexes ha are: he equally-weighed index and he value-weighed index. The graph for he Euro.NM shows a big difference in he resuls obained by hese wo indexes. In fac, if he adjusmen is made by equally-weighed index, he resuls are almos he same as hose obained by he differen reference porfolios. On he oher hand, he use of he value- weighed index shows a durable decline over he 36 monhs ha follow he sock marker lisings. Our resuls can be explained by a performance largely superior for he large companies o ha in he small-sized companies. The NASDAQ resuls are more or less he same using he wo indexes Resuls by using he Buy-and-Hold mehod The able 7 presens our resuls by using he buy-and-hold reurns mehod. This mehod makes i possible o ge he reurns obained from he invesor who acquire he shares from companies who makes an iniial public offering and which are reained wihin a ime scale of 6, 2, 8, 24, 30 and 36 monhs. We use several alernaives, hough of as normal reurns, o adjus he gross reurns: he equally-weighed marke index, he value-weighed marke index, he size porfolios, MTBV porfolios and he size and MTBV porfolios. Independen of he adjusmen facor, he resuls show he exisence of posiive abnormal reurns in he wo samples over a six monh ime scale. However, he difference beween he wo samples is more significan in he oher ime scales. We observe posiive abnormal reurns of.25% and 54.47% for he Euro.NM sample over a hree-year period and beween 6.8% and 86.3% for he NASDAQ sample. For he Euro.NM sample, if he adjusmen is made by he value-weighed index, we noe exreme resuls ha can be explained by a large variance beween he large companies reurns and hose of he small-sized companies. Table 7 also shows he wealh relaive raio ha describes he average raio of he reurns of IPOs and he reference porfolio reurns during he same period. This raio is calculaed according he following equaion: 6 Our resul corroboraes ha of Rier (99) who, using he same mehodological procedure, finds a cumulaed abnormal reurn of 29.3%. 2