Privatization versus Private Sector Initial Public Offerings in Poland

Transcription

1 Prvatzaton versus Prvate Sector Intal Publc Offerngs n Poland Wolfgang Aussenegg * Venna Unversty of Technology Department of Fnance Address: Favortenstraße 9-11, A-1040 Venna, Austra, Europe E-Mal: waussen@pop.tuwen.ac.at Phone: Fax: Frst draft: July 1999 Ths draft: May 2000 * I am grateful to Adam Gehr, Alos Geyer, Enrco Perott, Hans Retbauer, Peter Theodossou, Peter van Ojen, Davd Walker, partcpants of the 1999 Conference on Equty Market Development n Emergng and Transton Economes (Amsterdam, December 1999), the 2000 Mdwest Fnance Assocaton meetng (Chcago, Aprl 2000) and the 2000 Fnancal Management Assocaton European Conference (Ednburgh, Ma 2000) for helpful comments and suggestons. Fnancal support of the Venna Unversty of Technology n acqurng data used n ths study s gratefully acknowledged. The author would also lke to thank Reuters Ges.m.b.H (Venna) for provdng data.

2

3 Abstract Ths study compares the characterstcs and the prce behavor of case-by-case prvatzaton ntal publc offerngs and prvate sector ntal publc offerngs n Poland over the frst nne years after the reopenng of the Warsaw Stock Exchange n Aprl There s evdence that the Polsh government s market-orented, tryng to buld up reputaton for ts prvatzaton polcy over tme by underprcng, sellng a hgher fracton at the ntal offer and underprcng more when sellng to domestc retal nvestors. In the long run prvatzaton ntal publc offerngs experence nether an under- nor an overperformance. Hgh demand multples at the ntal offer result n a bad long-run performance, suggestng over-optmstc nvestors. Keywords: prvatzaton, ntal publc offerngs, Poland, underprcng, long-run performance JEL classfcaton Codes: G12; G18; G38

4 1. Introducton In the past two decades a lot of state enterprses have been prvatzed n the world. Ths apples to developed economes lke the UK, France or Germany as well as a lot of developng economes n Asa and South Amerca. In Central and Eastern European transton economes the gong publc process of state enterprses started only at the begnnng of the nnetes. After the breakdown of communsm a process of restructurng and transformaton from orgnally planned to market-orented economes was started. An mportant ngredent n such a radcal transformaton s the prvatzaton of state enterprses. Whle n western economes state enterprses are transferred nto a wellfunctonng economc system, ths s not the case n Central and Eastern European countres where a sutable legal and nsttutonal framework s not avalable at the start of the transformaton process. Ths study examnes the characterstcs as well as the short- and long-run prce behavor of Polsh case-by-case prvatzaton ntal publc offerngs (PIPOs) compared to Polsh prvate sector IPOs (IPOs). After a break of more than 50 years due to communsm the share tradng on the Warsaw Stock Exchange (WSE) resumed on Aprl 16 th, 1991 wth the lstng of fve companes. Snce then over 200 frms have gone publc on the WSE and more than 50 of these are prvatzed state-owned enterprses. In contrast to other Central and Eastern European countres, lke Russa, the Czech Republc or Slovaka, Poland decded to start the prvatzaton program wth case-by-case prvatzatons nstead of a mass prvatzaton program. 1 The followng questons are analyzed n ths study: Frst, can the Polsh government n the sense of Perott (1995) and Bas and Perott (1999) be classfed as market-orented and s a reducton n the poltcal uncertanty about the future prvatzaton polcy observable n the course of tme? Second, how do PIPOs and prvate sector IPOs dffer 1 A mass prvatzaton program was carred out n Poland only n a second phase of the prvatzaton process. At the end of 1999 ts sze (market captalzaton) reached only 1.5 percent of the sze of all frms lsted on the WSE. 1

5 wth respect to underprcng? Is the underprcng of PIPOs hgher than that of prvate sector IPOs? Thrd, how can the underprcng of PIPOs and prvate sector IPOs be explaned? Are explanaton hypotheses orgnally desgned for developed markets also vald for Polsh prvate sector IPOs and to what extent do they apply to PIPOs? Fourth, do Polsh PIPOs experence a postve long-run performance and s the long-run performance of prvate sector IPOs negatve as documented for many developed markets? Ffth, whch factors nfluence the long-run performance of Polsh PIPOs and prvate sector IPOs? Ths study contrbutes to the exstng lterature n several ways: Frst, an Eastern European emergng market s examned completely from the resumpton of the stock market. Other studes, lke Jones et al. (1999), Meggnson et al. (1999), Dewenter and Malatesta (1997) or Huang and Levch (1998) nvestgate several markets all over the world and consder developed and emergng markets jontly. The analyss of a sngle (Eastern European) emergng market can provde addtonal nsghts as the database s more homogeneous. Second, the nvestgaton of a share prvatzaton program rght from the begnnng allows e.g. testng to what extent a buld-up of reputaton for the government concernng the prvatzaton program takes place over tme. Thrd, n addton to PIPOs also prvate sector IPOs are examned and both groups are compared to each other. Ths allows dentfyng smlartes and dfferences between prvatzaton and prvate sector IPOs concernng ther characterstcs as well as ther short- and long-run prce performance. 2 The followng secton revews the emprcal evdence of studes examnng the prce performance of PIPOs. The thrd secton dscusses the theory and testable hypotheses about the short- and long-run prce behavor of PIPOs and prvate sector IPOs. Secton four descrbes the database used and provdes descrptve statstcs for the IPO groups. 2 Untl now there have only been a few comparsons of ths knd n the lterature. Examples are Menyah and Paudyal (1996), who examne prvatzaton and prvate sector IPOs for the UK, Paudyal et al. (1998), who compare these groups for Malaysa, and Jelc and Brston (1999), who analyse the prce behavor of Hungaran prvatzatons and prvate sector IPOs. The studes of Dewenter and Malatesta (1997) and Huang and Levch (1998) also compare prvatzaton and prvate sector IPOs but they only analyze the underprcng. 2

6 In the ffth secton the emprcal evdence regardng the ntal return s presented and possble reasons for the underprcng are examned. Secton sx deals wth the aftermarket performance and secton seven gves a summary of the fndngs. 2. Revew of the emprcal lterature on the short- and long-run prce behavor of prvatzaton ntal publc offerngs Numerous studes focusng on prvately-owned unseasoned new ssues document two man phenomena: Frst, frms gong publc are underprced and second, they tend to underperform benchmark frms n the long run. 3 Studes analysng the prce behavor of prvatzed state-owned enterprses show that PIPOs are, on average, also underprced. One of the frst studes n ths context was carred out by Jenknson and Mayer (1988). They fnd that French PIPOs are underprced by an average of 25.1% and PIPOs n the Unted Kngdom (UK) are underprced by an average of 22.2%. 4 In addton, Perott and Guney (1993) provde evdence that prvatzatons n Malaysa, Span and Turkey are underprced as well. Dewenter and Malatesta (1997) report for a total sample of 109 PIPOs of eght countres an (average) underprcng of 23.7%. For a subsample of 19 Polsh PIPOs they document a mean market-adjusted ntal return of 50.0%. In a recent paper Jones et al. (1999) examne a 59-country sample of PIPOs for whch they document an average ntal return of 34.1%. 5 Smlar results are provded by Huang and Levch (1998). A sgnfcantly postve underprcng can also be observed for Hungaran PIPOs (see Jelc and Brston, 1999). The results of nne studes examnng ntal returns are summarzed n table 1, Panel A See e.g. Loughran, Rtter, and Rydqvst (1994) for a survey revewng the nternatonal emprcal evdence. A non negatve long-run performance s e.g. documented for Korea (Km, Krnsky, and Lee (1995)), Sngapore (Lee, Taylor, and Walter (1996)) or Swtzerland (Kunz, and Aggarwal (1994)). Qualtatvely smlar fndngs are documented e.g. by Vckers, and Yarrow (1988), Levs (1993) or Menyah and Paudyal (1996). Ther 59-country sample conssts of 242 prvatzaton ntal publc offerngs (PIPOs), wth 26 PIPOs n the Polsh subsample. 3

7 *** Insert table 1 about here *** The long-run performance of PIPOs has only been examned n recent years, at least on a large scale. For a 33-country sample of prvatzatons Meggnson et al. (1999) fnd that regardless of the benchmark used the aftermarket performance of PIPOs s sgnfcantly postve. Menyal and Paudyal (1996) report smlar results for a sample of 40 UK PIPOs. But not all studes document a sgnfcantly postve long-run performance for PIPOs. Paudyal et al. (1998) fnd for Malaysa a negatve (but not sgnfcantly negatve) aftermarket performance over the frst three years. Usng a 26-country sample of 120 PIPOs, Boubark and Cosset (1999) report nsgnfcant market-adjusted average returns for a three-year holdng perod of 37% to 46% (Medan: -7% to 13%). Nne PIPOs n Chle experenced a sgnfcantly negatve one-year aftermarket performance (see Aggarwal et al., 1993). Panel B of table 1 provdes a summary for sx studes measurng the long-run performance of PIPOs. 6 Studes explctly comparng the characterstcs and the prce behavor of PIPOs and prvate sector IPOs are scarce. Dewenter and Malatesta (1997) test for dfferences between the mean ntal return n PIPOs and the mean ntal return n prvate sector IPOs. For a 7-country sample (Canada, France, Hungary, Japan, Malaysa, Thaland, and the UK) they fnd no general tendency for prvatzatons to be underprced more than prvate sector IPOs. Only n the UK do government offcals underprce publc share offers of state-owned frms sgnfcantly more than ther prvate company counterparts. For Canada and Malaysa, however, the opposte s true. In contrast, Paudyal et al. (1998) report that the average ntal return on Malaysan PIPOs s sgnfcantly hgher than on prvate sector IPOs. Jelc and Brston (1999) compare Hungaran PIPOs and prvate sector IPOs and report average ntal market adjusted returns of 44% and 40% respectvely. Consstent wth Dewenter and Malatesta (1997) they document no sgnfcant dfference between the ntal returns for Hungaran PIPOs and prvate sector IPOs. 6 Evdence for a sgnfcantly postve long-run performance of PIPOs s also provded by Cho et al. (2000) and Dewenter and Malatesta (2000). 4

8 Three studes explctly comparng the long-run performance of PIPOs and prvate sector IPOs are shown n table 2. The three-year buy-and-hold abnormal return for UK PIPOs, for nstance, s +61.0% for PIPOs and only +4.9% for prvate sector IPOs (see Menyal and Paudyal, 1996). For Hungaran ssues Jelc and Brston (1999) report a postve 3-year aftermarket performance for PIPOs but a negatve one for prvate sector IPOs. They show that PIPOs perform sgnfcantly better n the long run than IPOs. In contrast to these fndngs, Paudyal et al. (1998) document no abnormal long-run aftermarket performance dfference between prvatzatons and prvate sector IPOs n Malaysa. 7 *** Insert table 2 about here *** 3. Theores of short- and long-run prce behavor and testable hypotheses 3.1 Underprcng Many theores have been put forward to explan the underprcng of ntal publc offerngs. In order to formulate hypotheses to explan the frst-day return levels of Polsh PIPOs and prvate sector IPOs, the most common and, n the case of Poland, testable explanatons wll be examned. In Rock s model (1986) underprcng s a necessary equlbrum condton n a world of nformatonal asymmetry between groups of nformed and unnformed nvestors. Informed nvestors are able to dstngush between underprced ( good ) and overprced ( bad ) new ssues and therefore avod overprced IPOs. Good ssues are subscrbed for by nformed as well as unnformed nvestors whle bad ssues are subscrbed for by 7 An extensve survey of emprcal studes on prvatzatons are provded by Meggnson and Netter (2000). 5

9 unnformed nvestors only. It can therefore be expected that there s a postve relatonshp between the demand for an ssue and the underprcng level. 8 Another consequence of nformatonal asymmetry s that the level of underprcng requred to attract nvestors ncreases wth the ex-ante uncertanty about the value of the frm (see e.g. Beatty and Rtter (1986)). The greater the ex-ante uncertanty, the more costly t s to value a new ssue and the hgher s the expected reward nvestors demand. In order to test ths proposton t s necessary to use a proxy for the unobservable exante uncertanty. In the emprcal lterature t s common to measure the ex-ante uncertanty by the standard devaton of daly returns n the frst tradng months. 9 The asymmetrc nformaton theores also mply that the uncertanty about the value of small, not establshed frms s hgher than for offers of larger, well-known frms. Ths mples e.g. that offers of small, not establshed frms should be underprced more than offers of large, well-known frms. As PIPOs tend to be larger than prvate sector IPOs ths hypothess suggests that ntal returns n PIPOs should be lower than ntal returns n prvate sector IPOs. In another group of models underprcng s used as a sgnal that the frm s of a hghvalue type (see e.g. Allen and Faulhaber (1989), Grnblatt and Hwang (1989), and Welch, (1989)). These models assume that ssuers possess better nformaton about the true value of the offer than nvestors. In the sgnalng approach by Welch (1989), for example, hgh-qualty frms delberately choose an offer prce below the ntrnsc value to sgnal ther qualty to nvestors. Ths underprcng s motvated by the possblty of achevng hgher offer prces n subsequent seasoned ssues. Hgh-qualty frms wll therefore underprce ther ssue and only offer a small part of ther share captal at the ntal ssue. Ths mples a negatve relatonshp between the level of underprcng and the fracton of the share captal sold at the ntal offer. 8 9 Paudyal et al. (1998) e.g. document for Malaysan PIPOs as well as prvate sector IPOs a sgnfcantly postve assocaton between the demand multple and the underprcng. Smlar results for UK are presented by Menyah and Paudyal (1996). See for example Rtter (1984) or Paudyal et al. (1998). 6

10 A government commtted to prvatzaton has the objectve that publc offers of ts state enterprses should be a success. One requrement for success s that the market prce on the frst tradng days does not fall below the ssue prce. To keep the probablty of an unsuccsessfull ssue low, a hgher market volatlty wll encourage the ssung government to set a lower offerng prce, whch results n a hgher underprcng. It s therefore expected that the market volatlty pror to the ssue s n a postve relatonshp wth the underprcng. Followng Menyah and Paudyal (1996) and Paudyal et al. (1998), the market volatlty s measured n the present study usng the standard devaton of daly stock market ndex returns over three months pror to the subscrpton perod. Perott (1995) explctly models prvatzatons wthn a sgnalng framework. Smlar to Allen and Faulhaber (1989), Grnblatt and Hwang (1989) or Welch (1989), Perott (1995) presumes that there s nformatonal asymmetry between the ssuer n the case of prvatzatons the ssung government and the nvestors. An mportant feature of Perott s (1995) model s that the ssuer (the government) can nfluence a frm s value after the ntal offer. In hs model Perott (1995) dstngushes between two types of governments: marketorented (commtted) and populst governments. A market-orented government wants to prvatze state-owned enterprses serously and rreversbly. Ths does not apply to populst governments. Prvatzaton can only restran but not elmnate publc nterference, e.g. to transfer value from shareholders to other groups by polcy changes through e.g. regulaton or taxaton. A government commtted to prvatzaton wll not use such nstruments to transfer value, but less commtted or populst governments wll. Whether a government s market-orented or populst can only be realzed by the market f there are concrete sgnals. Only announcng to be market-orented s not enough. What sgnals can a government send to be recognzed as market-orented? In Perott s (1995) model a government has two parameters to sgnal: the fracton of the share captal t sells at the ntal offer and the ssue prce. A sellng government may sgnal ts commttment by retanng a stake n the frm for some tme, thus showng wllngness to 7

11 bear some fnancal costs of polcy changes. At the begnnng of the prvatzaton process the poltcal uncertanty can be expected to be hghest. At ths pont the porton a government retans from ts state-owned enterprses should therefore be hghest. On the other hand, f the poltcal uncertanty s very hgh, a commtted government mght be forced to sell a hgher fracton at the ntal offer, otherwse the market mght get the mpresson that the government possbly does not want to gve up control rghts over the state enterprse. If a government sells a large fracton n an envronment of hgh poltcal uncertanty, t transfers rsk to nvestors. To convnce the market to absorb a large porton and to sgnal commtment a government wll have to underprce. A testable mplcaton of Perott s (1995) model s therefore that the relatonshp between the fracton sold at the ntal offer and the underprcng level s postve for a market-orented government. For a market-orented government t can be expected that the poltcal uncertanty drops after the begnnng of the prvatzaton program due to the buldup of reputaton over tme. Ths has the testable mplcaton that for a commtted government the fracton of PIPOs sold at the ntal offer should be hghest at the begnnng of the prvatzaton program and drop thereafter. The same should be vald for the underprcng of PIPOs. By contrast, a populst government s not wllng to accept a (hgh) underprcng and the correspondng lower gross proceeds. A populst government would therefore also try to sell a not too small fracton, but at a maxmum ssue prce (and therefore mnmum underprcng). Bas and Perott (1999) show that a market-orented government can generate support for ts prvatzaton program by allocatng more shares to medan-class voters. Snce a reversal of the prvatzaton program decreases the value of the already prvatzed frms, medan-class voters wll support the prvatzaton efforts of the government, whch results n a support of the government n electons (re-electon hypothess). In order to make medan-class voters buy enough shares so that ther poltcal preferences are smlar to those of the government, underprcng s necessary. A government can acheve ths am by allocatng a large porton to domestc retal nvestors (medan-class voters) when sellng a large fracton of a state enterprse. If ths hypothess apples, the 8

12 underprcng should be hgher for ssues that meet two requrements: a) a relatvely large fracton s sold at the ntal offer and b) a relatvely large porton s allocated to domestc retal nvestors. The testable hypotheses to explan the underprcng of Polsh PIPOs and prvate sector IPOs can be summarzed as follows: Hypothess 1: Demand Multple: A hgher demand multple at the ntal offer s assocated wth a hgher underprcng. Hypothess 2: Ex-ante uncertanty I: The relatonshp between the level of underprcng and the ex-ante uncertanty measured by the standard devaton of the daly returns over the frst two aftermarket months s postve. Hypothess 3: Ex-ante uncertanty II: The underprcng level of PIPOs s lower than for prvate sector IPOs. Hypothess 4: Sgnalng: The relatonshp between the level of underprcng and the fracton of the share captal sold at the ntal offer s negatve. A hgher fracton sold s assocated wth a lower underprcng. Hypothess 5: Market Volatlty: The relatonshp between the level of underprcng and the standard devaton of daly returns of the Warsaw Stock Exchange (WIG) Index over three months pror to the subscrpton perod s expected to be postve. Hypothess 6: Poltcal Uncertanty I: The relatonshp between the level of underprcng and the fracton of the share captal sold at the ntal offer s postve for PIPOs Note that ths hypothess contradcts hypothess 4. A hgh uncertanty about the true value should lead to a negatve relatonshp between the level of underprcng and the fracton sold (hypothess 9

13 Hypothess 7: Poltcal Uncertanty II: If a large fracton of the share captal s sold at the begnnng of the prvatzaton process (when the poltcal uncertanty s expected to be hghest) to sgnal the wllngness to gve up control rghts, a commtted government has to underprce more. For such a government the underprcng s therefore expected to be hghest at the begnnng of the process and to drop (n lne wth the fracton sold) n the course of tme due to the buldup of reputaton. Hypothess 8: Re-Electon: Issues experence a hgher underprcng f the government sells a large fracton of the share captal at the ntal offer (.e. a fracton above the medan fracton sold) and f the government at the same tme allocates a large porton of the share captal to domestc retal nvestors (.e. a porton above the medan porton allocated). 3.2 Long-Run Performance The emprcal evdence of the long-run performance of frms gong publc ndcates that PIPOs and prvate sector IPOs do not perform smlarly. Prvate sector IPOs mostly experence a negatve abnormal performance over the frst three to fve years of aftermarket tradng, 11 whereas PIPOs mostly experence a long-run aftermarket performance equal or better than that of benchmark frms ). On the other hand, a hgher poltcal uncertanty mght requre sellng a larger fracton to transfer control rghts credbly, whch n turn forces a commtted government to underprce more, whch results n a postve relatonshp between the level of underprcng and the fracton sold (hypothess 6). See e.g. Rtter (1991) and Loughran and Rtter (1995) for US IPOs, Levs (1993) for UK IPOs or Keloharju (1993) for Fnnsh IPOs. Panel A of table 1 provdes evdence gven n several emprcal studes measurng the long-run performance of PIPOs. 10

14 A mostly non-negatve long-run abnormal performance for PIPOs concdes wth the objectves of a market-orented government. As prvatzaton programs n most cases last several years, a commtted government wll be nterested n buldng up reputaton for ts prvatzaton program over tme. Ths s the only way to generate support n the populaton, whch s necessary to successfully contnue the program. For Poland the followng hypothess s tested n ths context: Hypothess 9: For PIPOs the long-run aftermarket performance from the frst tradng day tll the thrd annversary s non-negatve. As there s (at least some) evdence that PIPOs tend to outperform prvate sector IPOs n the long-run (see e.g. the emprcal evdence n Menyah and Paudyal (1996) for the UK and Jelc and Brston (1999) for Hungary) t s also tested whether the long-run abnormal performance of Polsh PIPOs sgnfcantly exceeds that of Polsh prvate sector IPOs: Hypothess 10: The long-run abnormal performance of PIPOs from the frst tradng day tll the thrd annversary s sgnfcantly better than the 3-year abnormal performance of prvate sector IPOs. Boycko, Shlefer and Vshny (1996) show n ther model that the fracton a government sells at the ntal offer s an mportant factor for the restructurng efforts of state enterprses. The hgher the fracton sold, the lower s the possblty that poltcans nterfere drectly. Boycko, Shlefer and Vshny (1996) conclude that the relatonshp between effcent restructurng actvtes and the fracton of the state enterprse sold at the ntal offer should be postve. Provded that a lower state holdng leads to a more effcent restructurng, the long-run abnormal performance should be postve. In ths context the followng hypothess s tested: Hypothess 11: Poltcal Influence: The hgher the fracton of the share captal sold at the ntal offer, the lower s the drect poltcal nfluence. Ths mples 11

15 a better restructurng and therefore a better long-run abnormal performance. It s often cted that over-optmstc nvestors are one of the man reasons for a negatve long-run performance (see e.g. Aggarwal and Rvol (1990), Rtter (1991) or Loughran and Rtter (1995)). Frms plannng to go publc make use of ths over-optmsm and tme ther IPO correspondngly. If nvestors lose ther over-optmsm n the course of tme, ths leads to a bad long-run performance (nvestors sentment hypothess). The level of nvestors optmsm cannot be observed drectly. As a proxy the demand multple s used n ths study. 13 The two testable hypothess can be summarzed as follows: Hypothess 12: Investors sentment: Frms wth a hgh demand multple experence a bad long-run performance. As the cut-off pont between hgh and low demand multples the medan s used. 4. Data and sample characterstcs In Poland shares are traded on the Warsaw Stock Exchange (WSE), whch was founded n Durng the Second World War and the followng decades of communsm the WSE was closed. The WSE resumed tradng on Aprl 16 th, 1991 wth the lstng of fve frms. From Aprl 1991 tll the end of 1999 altogether 215 frms went publc wth a lstng on the WSE. 107 frms started tradng n the man market of the WSE, 78 n the parallel market and 30 n the free market. At the end of frms were lsted on the WSE: 119 n the man market, 61 n the parallel market and 26 n the free market. The dfferences to the number of gong publcs can be explaned by changes from one market segment to another and nne delstngs. 14 The parallel market started on Aprl 22 nd, 1993 and the free market on February 20 th, Ljungqvst (1996) provdes further varables that can be used as proxes for nvestors sentment. To avod a delstng bas all delsted frms are ncluded n ths study. 12

16 Ths study uses the two bggest market segments: the man market and the parallel market. They comprse the bggest and most lqud frms tradng on the WSE. Dsclosure requrements are the same for all companes, regardless of ther quotaton market. All companes are quoted n the sngle-prce aucton system. In addton, shares of the most lqud companes from the man and the parallel markets are traded n the contnuous tradng system. The man market and the parallel market dffer wth respect to the lstng requrements. A company whch would lke to begn tradng n the man market has to fulfll hgher mnmum sze and lqudty crtera than a company applyng for lstng on the parallel market. 15 The nvestgaton perod starts wth the reopenng of the WSE on Aprl 16 th, 1991 and ends on December 31 st, Of the 185 frms gong publc n the man market and the parallel market only those are used whch a) made a publc offer before tradng on the WSE and b) for whch all necessary nformaton was avalable (especally ssue prce, ssue volume, demand multple, contract type (fxed prce or tender), dvdend payment or seasoned equty offerng nformaton). Therefore the fnal sample comprses 159 frms, 99 from the man market and 60 from the parallel market. For nvestgaton purposes the sample of 159 frms s dvded nto two parts: A sample of 52 PIPOs and a sample of 107 prvate sector IPOs. 16 In the sample of PIPOs all those frms are ncluded whch are classfed as Publc Offerng by State Treasury n the WSE Fact Book These are frms controlled by the Polsh state and for most of them 100% of the shares were held by the state pror to gong publc Examples of lstng requrements (PLN numbers vald n the years 1998 and 1999) are the mnmum value of shares to be admtted for tradng (PLN 40 ml. for the man market versus PLN 14 ml. for the parallel market), the mnmum book value of the company (PLN 40 ml. versus PLN 14 ml.), the mnmum number of shareholders who hold shares to be admtted for tradng (500 versus 300) and the perod for whch the company s requred to dsclose audted fnancal reports to the publc (three fnancal years for the man market versus two fnancal years for the parallel market). Of the 52 PIPOs 50 went publc n the man market and 2 n the parallel market. Of the 107 prvate sector IPOs 49 went publc n the man market and 58 n the parallel market. 13

17 To buld up the database three man sources were used: Frst, the Hstorcal Database of the Warsaw Stock Exchange, whch provdes nformaton on daly stock prces, dvdend payments, number of outstandng shares and the frst tradng day on the WSE; second, several edtons of the WSE Fact Book whch were used to get nformaton about ssue prces, frst tradng day, orgnal ownershp structure of lsted companes, dvdend payments, stock splts, mergers and delstngs; and thrd, REUTERS Busness Brefng Archves 17 for nformaton on subscrpton perod, ssue prce, ownershp structure, demand multple, ratonng data, gross proceeds, seasoned equty offerngs, stock splts and other company-related nformaton. All nformaton receved from these three man sources was checked and completed usng nformaton avalable of the Polsh stock market n Reuters 3000 Equtes Hstory as well as n Datastream. Table 3 presents summary statstcs for the sample of all 159 ssues (All), the 52 PIPOs and the sample of 107 prvate sector IPOs (IPOs). About 84.5% of the total gross proceeds of 18,843 mllon PLN comes from the 52 PIPOs (15,932 Mllon PLN). The mean (medan) gross proceeds for the PIPO frms are more than 11 (3) tmes hgher than for the prvate sector IPOs (306.4 versus 27.2 mllon PLN and 43.5 versus 14.4 mllon PLN). The mean as well as the medan gross proceeds are sgnfcantly dfferent from each other (see table 3, Panel B). Ths has two reasons: Frst, PIPO frms are larger than prvate sector IPOs, and second, the fracton sold at the ntal offer s nearly twce as hgh for the PIPO frms as for prvate sector IPOs (63.2% versus 32.1%). The two samples sgnfcantly dffer from each other n ths respect. The mean percentage of captal offered n the 59-country sample of PIPOs n Jones et al. (1999) s 43.9%. Therefore, the Polsh government sold about 19 percentage ponts (or 45%) more than the average fracton sold on a global bass. 18 In Malaysa, for nstance, the government sold only 25% (see Paudyal et al. (1998)) and for Hungary the correspondng number s 41.7% (see Jelc and Brston (1999)) REUTERS Busness Brefng Archves s a comprehensve busness database, wth access to natonal and nternatonal news wres, news papers, trade journals, research reports and news pctures. For Non-UK PIPOs Jones et al. (1999) document an average fracton sold at the ntal offer of only 37.9%. Smlar results can be found n Huang and Levch (1998). 14

18 The model of Perott (1995) predcts that f there s hgher poltcal uncertanty about the future prvatzaton polcy, a market-orented government wll sell a hgher fracton at the ntal offer and wll underprce more to sgnal ts commtment. Ths mples that for a market-orented Polsh government the ntal return should be hgher than the global average (medan) of 34.1% (12.4%) provded n Jones et al. (1999). As secton fve shows ths apples to the 52 PIPO sample used n ths study. *** Insert table 3 about here *** PIPOs and prvate sector IPOs not only dffer wth respect to the gross proceeds and the fracton sold at the ntal offer but also sgnfcantly dependng on whether secondary or prmary shares are sold. 94.4% of the gross proceeds for PIPO frms consst of secondary shares and for two thrds of PIPOs only secondary shares were sold. In contrast, only prmary shares were sold at the ntal offer n 84.1% of prvate sector IPOs. Ths observaton s consstent wth the evdence provded by Meggnson, Nash, and Van Randenborgh (1994) and Jones et al. (1999), whch says that PIPOs n most cases tend to be pure secondary offerngs wth none of the sale proceeds flowng to the frm. On the other hand, prvate sector IPOs tend to be mostly prmary, captal-rasng events. As the evdence for Poland shows, ths seems to be a general feature of prvate sector IPOs. The emprcal evdence for Poland reveals that only 57.7% of all PIPOs n the sample have a fxed ssue prce. Ths compares to the global evdence n Jones et al. (1999) of 79.9% fxed-prce offers for ther 59-country sample. Ths ndcates that the Polsh government dd not excessvely need fxed-prce offers to generate enough demand for ts ssues. The average demand measured by the demand multple s only a lttle hgher for PIPOs than for the prvate sector IPOs (2.63 versus 1.96). 19 But both mean and medan demand multples of the two samples are not sgnfcantly dfferent from each other. 19 Much hgher demand multples are reported for Malaysa, wth mean demand multples of 21.2 for PIPOs and 23.4 for prvate sector IPOs (see Paudyal et al. (1998)). 15

19 In Poland trade unons have hstorcally been very powerful. In the process of prvatzaton t can therefore be expected that the government wll partcularly try to offer shares to the employees of the state enterprses. On average (medan) 19.0% (18.7%) of the shares sold durng the ntal offer were sold to employees. Ths value s more than twce as hgh as the nternatonal evdence n Jones et al. (1999). For ther 59-country sample they report an average (medan) porton sold to employees of 8.5% (7.0%). 20 Ths underlnes the strong and mportant poston of trade unons n Poland and the poltcal necessty to ntegrate Polsh employees n the prvatzaton program to a hgher degree than n most other countres. 5. Intal return 5.1 Methodology and summary statstcs The startng pont for nvestgatng the ntal returns s the calculaton of ntal raw returns and ntal market-adjusted returns. The ntal raw return for IPO corresponds to ts buy-and-hold return (BHR) from the ssue prce to the closng prce on the frst tradng day and s defned as BHR P,1 P,0 =, (1) P,0 where P,0 represents the ssue prce and P,1 the closng prce on the frst tradng day of IPO. The tme ndex t=0 refers to the frst day of the subscrpton perod. The marketadjusted return (=underprcng) for each ssue s defned as the dfference between ts ntal raw return (the BHR ) and the correspondng return on the market ndex: UP = BHR BHR, (2) WIG, 20 For an nternatonal sample of 190 PIPOs Huang and Levch (1998) document an average porton sold to employees of 9.5%. 16

20 where UP s the underprcng of IPO and BHR WIG, s the buy-and-hold return of the Warsaw Stock Exchange Index (WIG). 21 Analogous to (1) the BHR for the WIG-Index s defned as BHR WIG, WIG,1 WIG,0 =. (3) WIG,0 WIG,0 represents the WIG-Index on the frst day of the subscrpton perod of IPO and WIG,1 s the WIG-Index at the end of the frst tradng day of IPO. Panel A of table 4 presents summary statstcs of the raw and market-adjusted ntal returns for all three samples. Wth the tradng strategy of nvestng a smlar amount of money n each of the 159 Polsh IPOs at the ssue prce and sellng each IPO on ts frst tradng day an nvestor would have earned an average ntal raw return of +38.5%. Appled to the ndvdual subsamples ths tradng strategy would have yelded average ntal raw returns of +65.6% and +25.3% for the PIPOs and the prvate sector IPOs respectvely. The market-adjusted average ntal returns are also all postve: 60.4% for PIPOs, 19.8% for prvate sector IPOs and 33.1% for the sample of all 159 ssues. All mean and medan returns reported n Panel A of table 4 are greater than zero at the 1% sgnfcance level. In addton, about a quarter of all Polsh IPOs are overprced (negatve ntal market-adjusted return). These results document that, smlar to the fndngs reported n earler research for other markets, Polsh PIPOs as well as Polsh prvate sector IPOs are sgnfcantly underprced. *** Insert table 4 about here *** For a 59-country sample of 303 PIPOs Jones et al. (1999) present an average (medan) ntal raw return of 34.1% (12.4%). The average ntal raw return for Polsh PIPOs s therefore nearly twce as hgh. Only two studes examnng ndvdual countres report a hgher underprcng for PIPOs compared to Poland: Paudyal et al. (1998) for Malaysa (103.5%) and Su and Flesher (1999) for Chna (948.6%). A lower underprcng s 21 The WIG s a value-weghted share prce ndex comprsng all frms lsted on the man market of the Warsaw Stock Exchange. 17

21 documented e.g. for Chle (7.6%, see Aggarwal et al. (1993)), for Hungary (44%, see Jelc and Brston (1999)), for the UK (38.7%, see Menyah and Paudyal, 1998) and France (25.1%, Jenknson and Mayer (1988)). Further detals of these studes are reported n table 1. The observaton that the Polsh government sold a hgher fracton at the ntal offer than the average reported for Jones 59-country sample s consstent wth the predctons n Perott (1995). A market-orented government sellng a hgher fracton at the ntal offer has to underprce more to sgnal ts comm tment. The mean raw and market-adjusted ntal returns of PIPOs are about 40 percentage ponts above those of prvate sector IPOs (see table 4, Panel B). But both test statstcs of and are statstcally nsgnfcant. The dfferences n the medan values are only +4.1% and +6.9%. The man reason for the large mean underprcng of PIPOs s one ssue: Bank Slask. After an ssue prce of 50 PLN the market prce on the frst tradng day rose to 675 PLN. The resultng ntal raw return of +1,250% caused a lot of crtcsm for the prvatzaton polcy and also poltcal consequences. 22 Wthout Bank Slask the average (medan) market-adjusted ntal return of the remanng 50 PIPOs drops to % (+19.74%), resultng n a decreased and statstcally nsgnfcant average (medan) dfference between PIPOs and prvate sector IPOs of 19.04% (6.83%). Altogether ths ndcates that the Polsh government does not sgnfcantly underprce ntal offers more than prvate company ssuers do. Hypothess 3, whch mples that the underprcng level of PIPOs s hgher than for prvate sector IPOs therefore has to be rejected. Accordng to hypothess 7, a commtted government has to underprce more to sgnal the wllngness to gve up control rghts f a large fracton of the share captal s sold at 22 The unusually hgh ntal return of Bank Slask led to the resgnaton of the deputy fnance mnster Stefan Kawalec, who was n charge of prvatzng the Polsh bankng sector. He was accused of havng underestmated the value of the stock resultng n a hgh loss of money for the state budget. Besdes the probable undervaluaton, there was also a second reason for the extremely hgh market prces on the frst tradng day: a delay n openng accounts and confrmng the purchase of the bank s shares. As a result, two thrds of all new shareholders (about 800,000 Polsh ctzens bought Bank Slask shares, only three shares were allocated per person) could not sell ther shares on the frst tradng day, whch led to a shortage of supply. 18

22 the begnnng of the prvatzaton program (when the poltcal uncertanty s expected to be hghest). Assumng that the poltcal uncertanty was hgh at the start of the program n Poland, the fracton sold and the underprcng should have been hghest at the begnnng of the program and should have decreased over tme due to the buld up of reputaton. To analyze these hypotheses the nvestgaton perod s dvded nto four subperods: 1991 tll 1992 (subperod one), 1993 tll 1994 (subperod two), 1995 tll 1996 (subperod three) and fnally 1997 tll 1999 (subperod four). Table 5 reports the emprcal results. Frst, as expected, the fracton of captal sold at the ntal offer s hghest for PIPOs at the begnnng of the program and decreases n the course of tme (see Panel A of table 5). The average (medan) fracton sold s 86.1% (90.0%) n the frst subperod, 63.6% (68.3%) n the second subperod, 54.6% (60.1%) n the thrd subperod and 48.0% (44.2%) n the last subperod. Second, the underprcng level of PIPOs s not hghest n the frst but n the second subperod and decreases thereafter (see Panel B of table 5). The average (medan) underprcng for PIPOs s 11.8% (13.1%) n the frst subperod, 191.0% (71.4%) n the second subperod, 46.8% (29.2%) n the thrd subperod and 15.8% (19.7%) n the last subperod. *** Insert table 5 about here *** The man reason for the hgh underprcng level n the second subperod ( ) s agan Bank Slask wth a huge underprcng of %. Wthout Bank Slask the average (medan) underprcng level of PIPOs drops to % (20.56%). But stll the average underprcng decreases after the second subperod. There are sx PIPOs underprced by more than 100% n the second subperod ( ) and two PIPOs underprced by more than 100% n the thrd subperod ( ). In the frst and fourth subperod no ssue s underprced by more than 100%. The low underprcng and the hgh fracton sold n the frst subperod would suggest a populst (or at least a less commtted) government. But another nterpretaton s that the frst subperod was used as a test phase by the Polsh government. Ths vew s supported by the fact that state-owned enterprses sold n the frst subperod are very small 19

23 compared to PIPOs n the remanng perods. The average market values of PIPOs (measured usng the frst market prce) n the last three subperods are 14 to 182 tmes hgher than n the frst subperod. The underprcng level for prvate sector IPOs and the fracton sold at the ntal offer do not behave n the same way as for PIPOs. Throughout the perod consdered the fracton sold s nearly constant at around 30% and the underprcng level decreases but from a lower level than for PIPOs. All n all, the evdence n table 5 provdes no clear support for the vew of a commtted government n the sense of Perott (1995). Only f the frst subperod s excluded, do the fracton sold at the ntal offer and the underprcng behave as predced n hypothess Multvarate cross-sectonal analyss In ths secton, the determnants of underprcng n Polsh PIPOs and Polsh prvate sector IPOs are studed n a multvarate cross-sectonal analyss, testng hypotheses specfed n secton three. The followng regresson models are used: Model I: UP = α o + α 1 MV + α 2 DM + α 3 Rsk + α 4 LFR + ε (4.1) Model II: UP = α o + α 1 MV + α 2 DM + α 3 Rsk + α 4 Frac + ε (4.2) where UP s the underprcng (market-adjusted ntal return) of ssue, MV s the market volatlty measured usng the standard devaton of daly returns of the Warsaw stock exchange (WIG) ndex durng the last three months before the subscrpton perod of ssue (hypothess 5: market volatlty), DM s the demand multple of ssue (hypothess 1: demand multple), Rsk s the two month aftermarket daly return standard devaton of ssue (hypothess 2: ex-ante uncertanty I), LFR st a dummy varable: coded 1 f a large fracton of the share captal s sold at the ntal offer (.e. a fracton above the medan) and f n addton domestc retal nvestors receve a large porton of the ssue (.e. a porton above the medan) and 0 otherwse) (hypothess 8: re-electon) 20

24 and Frac s the fracton of the share captal sold at the ntal offer (hypothess 4: sgnalng and hypothess 6: poltcal uncertanty I). The varables LFR and Frac are hghly correlated and cannot be used n the same model. To test all hypotheses specfed above t s therefore necessary to use two models. Table 6 presents the regresson results. 23 Frst, the varable market volatlty (MV) s sgnfcantly postvely related to the underprcng level of Polsh PIPOs. In lne wth hypothess 5 ths result s consstent wth a commtted government tryng to make ts ssues a success n order buld up reputaton for further ssues of state enterprses. If the market condtons are rsker (hgher volatlty), t reduces the ssue prce to ensure the success of the ssue. Ths fndng for Polsh PIPOs s consstent wth the results for UK PIPOs reported n Menyah and Paudyal (1996) and for Malaysan PIPOs reported n Paudyal et al. (1998). In contrast to PIPOs there s no sgnfcant relatonshp between the underprcng level and the market volatlty observable for prvate sector IPOs. 24 *** Insert table 6 about here *** Second, the demand multple (DM) s sgnfcantly postvely related to the underprcng level. The postve relatonshp suggests that demand pressure pushes the market prce on the frst day upwards and results n a hgher market-adjusted ntal return. Ths fndng s consstent wth the results of Menyah and Paudyal (1996) for UK PIPOs and Paudyal et al. (1998) for Malaysan PIPOs. They both document a sgnfcantly postve relatonshp between the underprcng of PIPOs as well as prvate sector IPOs and the demand multple. How can ths result be nterpreted? A postve relatonshp between the ntal marketadjusted return and the demand multple ndcates that good ssues are subscrbed for by more nvestors, ncludng nformed nvestors. Ths would support Rock s (1986) wnner s curse model. If small (domestc) retal nvestors belong to the group of unnformed nvestors and nsttutonal nvestors to the group of nformed nvestors then the Bank Slask s vewed as an outler and s therefore not ncluded n the regresson analyss. Paudyal et al. (1998) document a smlar result for ther sample of Malaysan prvate sector IPOs. 21

25 demand of small nvestors should be ndependent of whether an ssue s over- or underprced. But ths s not the case n Poland. The demand multple of small retal nvestors for underprced ssues (postve market-adjusted ntal return) s 3.08 and 0.88 for overprced ssues (negatve market-adjusted ntal return). For nsttutonal nvestors the correspondng demand multples are 2.71 and 0.87 respectvely. 25 Ths ndcates that both groups of nvestors are able to dfferentate between good and bad ssues. Ths observaton s not n lne wth the wnner s curse model. Thrd, the underprcng level of Polsh PIPOs and prvate sector IPOs s unrelated to the varable Rsk, whch has a negatve nstead of an expected postve sgn. For PIPOs ths result s consstent wth the fndngs for Malaysan PIPOs (see Paudyal et al. (1998)). But for prvate sector IPOs the ex-ante uncertanty has a sgnfcant explanatory power n many markets. 26 Fourth, the results n table 6 reveal that the underprcng for PIPOs s especally large f a large fracton s sold at the ntal offer and f, n addton, domestc retal nvestors are allocated a large porton at the ntal offer. Ths result suggests that f the Polsh government wants to ncrease the number of shareholders and therefore also the acceptance of ts prvatzaton program among the ctzens, t sells a hgher fracton and uses an underprcng strategy. Ths emprcal evdence s n accordance wth the model of Bas and Perott (1999) and hypothess 8 can be accepted. By contrast, the varable LFR has no explanatory power for prvate sector IPOs. The fnal varable s the fracton of the share captal sold at the ntal offer. Hypothess 4 predcts a negatve relatonshp between the fracton sold and the underprcng level. But for PIPOs as well as prvate sector IPOs the relatonshp s not sgnfcantly negatve but postve, n the case of prvate sector IPOs even sgnfcantly postve. As the relatonshp for PIPOs s not sgnfcant, hypothess 6 has to be rejected too Informaton on the demand multple for small retal nvestors and nsttutonal nvestors s avalable for a subsample of 135 frms. See e.g. Rtter (1984) for US IPOs, Fnn and Hgham (1988) for Australan IPOs and Paudyal et al. (1998) for Malaysan prvate sector IPOs. 22

26 Other studes relatng the fracton of the share captal sold to the underprcng level of PIPOs were made by Jones et al. (1999) for an nternatonal sample of 93 PIPOs, Menyah and Paudyal (1996) for the UK and Paudyal et al. (1998) for Malaysan PIPOs. They report dfferent fndngs. Paudyal et al. (1998) and Jones et al. (1999) document that a hgher fracton sold s (sgnfcantly) related to hgher underprcng levels. By contrast, Menyah and Paudyal (1996) fnd a sgnfcantly negatve relatonshp for UK PIPOs. To sum up, three out of fve varables have a sgnfcant and postve mpact on the underprcng of Polsh PIPOs: the market volatlty pror to the ssue, the demand multple and a dummy varable for the re-electon hypothess of Bas and Perott (1999). The sgnalng hypothess (hypothess 4) has to be rejected and the postve relatonshp between the underprcng level and the fracton sold s not sgnfcant. On the other hand, only the demand multple and the fracton sold have sgnfcantly postve explanatory power for the underprcng of prvate sector IPOs. 6. Aftermarket Performance The objectve of ths secton s to examne the aftermarket performance of frms gong publc on the WSE. Besdes the short-run aftermarket performance, whch permts conclusons about the prce adjustment process, the long-run aftermarket performance wthn three years of the frst tradng day s analyzed. Four hypothess, specfed n Secton 3, are explctly tested: hypothess 9 (no negatve long-run performance for PIPOs), hypothess 10 (PIPOs experence a better long-run performance than prvate sector IPOs), hypothess 11 (a lower drect poltcal nfluence s assocated wth a better long run performance) and hypothess 12 (nvestors sentment: a hgh demand multple at the ssue leads to a bad long-run performance). 23

27 6.1 Methodology To measure the performance of IPOs n the aftermarket buy-and-hold returns are calculated n a frst step for each ssue. In contrast to cumulatve returns, whch are sometmes used to measure long-horzon securty, prce performance buy-and-hold returns have the advantage that they are based on a realstc ex-ante tradng strategy. The buyand-hold return for ssue (BHR,T ) s defned as BHR,T T = + t= 2 ( 1 R ) 1,t, (5) where R,t s the return of IPO n perod t and t = 2 ndcates the second tradng day n the aftermarket. 27 BHRs are calculated for the followng tme perods: T = 1 week, 2 weeks, 1 month, 2 months, 1 year, 2 years and 3 years. To be able to compare the aftermarket performance wthn the frst three years after the frst tradng day for dfferent tme perods (e.g. one year and three years) only ssues are used wth a frst tradng day earler than March 10 th, Ths leads to a reducton n the sample szes to 83 frms for the sample of all ssues, 38 frms for the sample of PIPOs and 45 frms for the sample of prvate sector IPOs. To measure the abnormal performance of IPOs n the aftermarket t s frst necessary to specfy approprate benchmarks. Ths task s of partcular mportance because t can affect the aftermarket performance measured. One possblty s to use a matchng frm adjustment procedure, n whch for each IPO frm a non-ipo frm of approxmately smlar sze and the same ndustry s chosen. 28 Another possblty, whch s used for markets n whch the number of potental benchmark frms s low, s to use one or more The startng pont for measurng the aftermarket performance s therefore the closng prce on the frst tradng day. See e.g. Rtter (1991) or Loughran and Rtter (1995). 24

28 ndces, e.g. a market ndex, as benchmark. 29 Because of the low number of frms lsted n the frst years after the resumpton of the WSE a matchng procedure (e.g. matchng by sze and ndustry) s not possble n Poland. The aftermarket performance s therefore measured aganst the WIG-Index as a benchmark. For comparson purposes and as a robustness check an equally weghted stock market ndex s used as an addtonal benchmark. 30 In a smlar way to (5) the BHR of the WIG-Index for IPO (BHR,WIG,T ) s calculated as BHR WIG,,T T = + t= 2 ( 1 R ) 1 WIG,,t. (6) R,WIG,t s the return of the WIG-Index n perod t, where t = 2 ndcates the second tradng day n the aftermarket. Therefore, buy-and-hold returns over dentcal ntervals are calculated for each ssue and the benchmark. To measure the market-adjusted performance buy-and-hold abnormal returns (BHARs) and wealth relatves (WRs) are used. In accordance wth Rtter (1991) the WR of IPO (WR ) s defned as WR 1+ BHR = (7) 1+ BHR WIG, and BHARs are defned as BHAR,T = BHR BHR. (8),T WIG,,T See e.g. Keloharju (1993) for the Fnnsh IPO market, Kunz and Aggarwal (1994) for the Swss IPO market, Paudayal et al. (1998) for Malaysan and Jelc and Brston (1999) for Hungaran PIPOs and prvate sector IPOs. It s common n the lterature to use value-weghted as well as equally weghted stock market ndces as benchmarks (see e.g. Loughran and Rtter (1995) or Brav and Gompers (1997)). The WIG Index s a value-weghted ndex. Small PIPOs or prvate sector IPOs are therefore prmarly compared wth bg frms. Ths s not the case when an equally weghted ndex used. 25