Delberate premarket underprcng and aftermarket msprcng: New nsghts on IPO prcng Abstract We decompose ntal returns nto delberate premarket underprcng and aftermarket msprcng usng Stochastc Fronter Analyss. We model delberate underprcng as a functon of proxes of nformaton asymmetry surroundng IPO value between market partcpants. Equty retaned s an unlkely sgnallng mechansm to convey IPO value to outsde nvestors through delberate premarket underprcng. The presence of lock-n agreements, underwrter fees, number of uses of proceeds, and venture captal or prvate equty backng have postve mpacts on delberate premarket underprcng. Demand for frms' captal also explans delberate premarket underprcng, whereas new ssues market condtons have no mpact. All these factors are found to explan a sgnfcant fracton of the varatons n our delberate underprcng estmates. Delberate underprcng s the more domnant component that makes up ntal return when compared to the fracton of aftermarket msprcng. We attrbute aftermarket msprcng to tradng volume n IPO shares on the frst day, prce adjustment between the flng prce range and the offer prce, and offer sze. Equty retaned explans the aftermarket msprcng rather than the delberate premarket underprcng n contradcton to the sgnallng argument. More reputable underwrters are lkely to provde prce support n the early aftermarket, whereas we observe no mpact on delberate premarket underprcng. Keywords: Intal IPO return, delberate premarket underprcng, aftermarket msprcng JEL: G1, G3, G3 1
1980q1 1981q1 198q1 1983q1 1984q1 1985q1 1986q1 1987q1 1988q1 1989q1 1990q1 1991q1 199q1 1993q1 1994q1 1995q1 1996q1 1997q1 1998q1 1999q1 000q1 001q1 00q1 003q1 004q1 005q1 006q1 007q1 008q1 009q1 010q1 011q1 01q1 013q1 Number of IPOs Average Intal Returns (%) 1. Introducton An ntal publc offerng (IPO) s the frst sale of shares by a prvate company to the publc on a securtes exchange. The valuaton of shares of an unlsted company s challengng due to the absence of current market prces and tradng hstory. Informaton asymmetry between market partcpants surroundng the value of these shares leaves the new ssues market subject to Akerlof s (1970) classc adverse selecton problem whch manfests tself n persstent average ntal return across captal markets. 1 Fgure 1 documents average ntal return and the persstence of the phenomenon as well as the number of U.S. IPOs by quarterly ntervals from the frst quarter of 1980 to the last quarter of 01. The dagram llustrates the cyclcty n persstent ntal return as well as the number of IPOs durng the sample perod. Fgure 1: Average ntal return and number of US IPOs by quarterly ntervals, 198001 Average ntal return s the equally-weghted mean of the percentage change from the offer prce n the premarket to the aftermarket prce. Quarterly data on ntal return and the number of IPOs are compled from monthly observatons avalable from Jay Rtter's webste (http://bear.warrngton.ufl.edu/rtter). 300 80 60 40 0 00 180 160 140 10 100 80 60 40 0 0 100 90 80 70 60 50 40 30 0 10 0-10 Year-Quarter Number of IPOs Average Intal Returns (%) 1 Intal return s the percentage change from the offer prce to the market prce n the aftermarket. See, for example, Table 1 n Banerjee, Da, and Keshab (011), Table 6 n Rtter (003), and Jay Rtter s web ste (http://bear.warrngton.ufl.edu/rtter) that has regular updates of Table 1 n Loughran, Rtter, and Rydqvst (1994). Ths s the URL address for all future references to Jay Rtter s IPO webste.
The lterature offers three man explanatons for persstent average ntal return: delberate underprcng n the premarket (henceforth delberate underprcng), msprcng n the early aftermarket as a result of tradng actvty, and underwrter prce stablzaton n the early aftermarket. 3 The frst strand of the lterature ascrbes ntal return to delberate underprcng as the outcome of nformaton asymmetry surroundng IPO value among partcpants n the premarket. For example, Grnblatt and Hwang (1989), Welch (1989), and Allen and Faulhaber (1989) use delberate underprcng as a costly and dffcult to mtate sgnal n whch frm nsders transmt postve nformaton on frm value to outsde nvestors. Baron (198) explans delberate underprcng and the persstence of the phenomenon as the outcome of a prncpal-agent conflct n whch the ssung frm (prncpal) cannot drectly observe the marketng and dstrbuton efforts of the underwrter (agent). The underwrter can thus nduce the ssung frm to agree to a relatvely low offer prce. Rock s (1986) model assumes that IPOs have to be, on average, underprced to compensate less-well nformed nvestors for a wnner s curse 4 problem and to nduce them to partcpate n the new ssues market. The second strand of the lterature assumes that IPOs are prced at ther ntrnsc value n the premarket and attrbutes ntal return to tradng actvty n the early aftermarket as a result of, for example, overoptmstc nvestors and ther valuatons. For example, Aggarwal and Rvol (1990) focus on fads n the IPO market whereby new ssues are possbly not prced at ntrnsc values n early aftermarket tradng. Rtter (1991) and Loughran and Rtter (1995) explan ntal return as a consequence of nvestors beng overoptmstc about frm value whch creates excess demand n shares, pushes up prces and leads to hgh ntal return n the aftermarket. Fresen and Swft (009) attrbute ntal return to nvestor overreacton at the tme of the IPO before prces revert back to fundamental frm value. Aggarwal (000, 003), Ells, Mchaely, and O Hara (000, 00) and Ells (006) fnd evdence that flppng s not solely responsble for hgh tradng volumes n the early aftermarket. 77% of trades are nvestor-motvated, whle nterdealer tradng accounts for the remanng 3%. Chahne (007) report a postve relatonshp between the share demand-to-offer rato n the premarket and the tradng volume n the aftermarket. The thrd strand of the lterature attrbutes postve average ntal return to underwrter prce support. The prce support leads to a censorng of the return dstrbuton and the spurous mpresson of persstent average ntal return. For example, Ruud (1993), Asquth, Jones and Keschnck (1998) and Aggarwal (000, 003) fnd that underwrters stablze the aftermarket prces at the offer prce whch results n very few IPO stocks beng overprced. 3 Premarket ndcates the tme leadng up to the IPO date (stock market flotaton), whereas aftermarket refers to the tme once tradng commences n the stock market. 4 A wnner s curse arses here because less-well nformed nvestors obtan a hgher proporton of overprced IPOs because better-nformed nvestors only apply for underprced new ssues. 3
On the one hand, exstng research explans postve average ntal return and ts persstence as a combnaton of delberate underprcng and aftermarket msprcng as a result of tradng n the early aftermarket. The most commonly held vew s that delberate underprcng s the man contrbutor towards persstent average ntal return (Grnblatt and Hwang, 1989; Welch, 1989; Allen and Faulhaber, 1989; Baron, 198; and Rock, 1986). On the other hand, emprcal studes do not dstngush between delberate underprcng and aftermarket msprcng when testng competng theores. 5 However, the lack of such a dstncton s not surprsng. Dsentanglng delberate underprcng and aftermarket msprcng from ntal return s dffcult. To do so requres a 'far' offer prce, whch s not drectly observable, n addton to the offer prce and the closng share prce n the aftermarket. In ths paper, we dentfy ths far offer prce usng stochastc fronter analyss (henceforth SFA), whch estmates the maxmum achevable offer prce from gven ssung frm attrbutes, deal characterstcs, thrd-party certfcaton, and IPO market condtons. SFA assumes the exstence of an effcent prcng fronter such that the offer prce falls onto or below the estmated fronter due to a systematc one-sded error to account for prcng neffcency plus a random error term. The one-sded error term allows calculatng an neffcency measure to uncover delberate underprcng from the maxmum offer prce that les on the effcent fronter. Our approach bulds on exstng studes n ths area, such as for example, Hunt-McCool, Koh and Francs (1996), Koop and L (001), Francs and Hasan (001), Chen, Hung and Wu (00), and Chan, Wu and Kwok (007). Compared to these earler studes, our stochastc IPO prcng fronter uses an extended set of prcng factors to estmate the maxmum (far) achevable offer prce. These prcng factors are the prmary drvers of equty value and orgnate from standard fnance theory. They nclude, for example, proxy varables for expected cash flows, cost of captal, and future growth prospects. These value drvers determne the effcent fronter of far offer prces. An ncomplete set of prcng factors could lead to less accurate estmates of far offer prces and hence less precse estmates of delberate underprcng. Our estmaton of the prcng fronter smultaneously controls for what we call underprcng factors to explan varatons n delberate underprcng. For ths, we use a set of varables drawng from exstng theores on nformaton asymmetry surroundng IPO value among new ssues market partcpants such as, for example, Grnblatt and Hwang (1989), Welch (1989), Allen and Faulhaber (1989), Baron (198), and Rock (1986). Fnally, after havng dentfed delberate underprcng, we uncover aftermarket msprcng, whch allows us to explan aftermarket msprcng wth the help of proxy measures that capture tradng actvty n the early aftermarket. We argue that IPO market value partly depends on ths tradng actvty and use relevant proxy varables from earler studes such as, for example, Chen, Hung and Wu (00), Ells (006), and Agarwal (003). 5 For an overvew of the emprcal lterature see, for example, Jenkngson and Ljungqvst (001). 4
We hence make several dstnct contrbutons to the lterature n lght of the ongong debate on whether delberate underprcng or aftermarket msprcng s the domnant explanaton of observed ntal return. Frst, we add to the lterature on the use of SFA as a tool for dfferentatng delberate premarket underprcng from ntal return. The use of SFA to measure delberate premarket underprcng n U.S. IPOs was frst proposed by Hunt-McCool, Koh and Francs (1996), wth later applcatons by Koop and L (001) and Francs and Hasan (001). 6 By estmatng delberate underprcng, we can test whether ntal returns s prmarly drven by delberate underprcng or by aftermarket msprcng as a result of early tradng n the aftermarket. Second, usng a more refned measure, we test the commonly held vew that nformaton asymmetry surroundng IPO value among new ssues market partcpants explans varatons n delberate underprcng. Whle exstng studes have tested the explanatory power of nformaton asymmetry theores, these have been conducted usng ntal return, whch s a combnaton of delberate underprcng and aftermarket msprcng. By constructng a more refned measure of delberate premarket underprcng, we are able to drectly test how t s affected by varables proxyng nformaton asymmetry, and hence mprove on these earler studes. We test the effect of nformaton asymmetry on delberate underprcng by explotng developments n the SFA lterature on the modellng of exogenous factors affectng the dstance from the fronter. For ths, we use the condtonal varance model proposed n Kumbhakar and Lovell (000), smultaneously estmatng the stochastc prcng fronter alongsde the determnants of delberate underprcng. Fnally, once we have estmated the delberate underprcng component, we recover an estmate of aftermarket msprcng and then test whether t can be explaned prmarly by tradng actvty. Accordngly, we regress ths aftermarket msprcng component of ntal return aganst proxes of tradng actvty, allowng us to test the hypothess that IPO market value partly depends on the tradng actvty n the early aftermarket.. Dsentanglng delberate premarket underprcng and aftermarket msprcng The central contrbuton of ths paper s to dsentangle delberate underprcng n the premarket and aftermarket msprcng from ntal returns, and test whether varatons n these components can be explaned by exstng theores. Dfferentatng these components from observed ntal returns reles crucally on beng able to estmate a far offer prce for each IPO. We estmate the far offer prce usng SFA, a benchmarkng econometrc technque that has seen numerous applcatons snce t was orgnally proposed ndependently by Agner, Lovell and Schmdt (1977) and Meeusen, and van de Broeck (1977). A few studes have used SFA as a tool to measure delberate underprcng n IPOs. Hunt-McCool, Koh and Francs (1996) s the frst study to apply SFA n IPO prcng. They use frm characterstcs, ndustry dummy varables and market sentment to estmate 6 To the best of our knowledge these are the only studes usng U.S. data. 5
the far offer prce and delberate underprcng on a sample of 1,035 US IPOs between 1975 and 1984. In ther study, offer prce s related to a number of prcng factors, ncludng book value, sales, number of rsks dsclosed n the prospectus, underwrter commsson rates, equty retaned by pre- IPO owners, frm age, offer proceeds from ssue, and prce-earnngs rato. The analyss dvdes the sample nto hot ssue and non-hot ssue perods to capture market sentment, whereby tme ntervals wth relatvely hgh ntal return and large numbers of IPOs are deemed to be hot new ssues markets. The authors estmate delberate underprcng at 8.9% and 8.0% respectvely for both hot-ssue and non-hot ssue markets. These estmates compare to observed average ntal return of 10.8% and 9.1% respectvely. In a further analyss, the authors regress ntal return (based on the frst avalable aftermarket bd prce) aganst the estmated delberate underprcng component and fnd a sgnfcant postve correlaton only for the hot ssues perod. They conclude that delberate underprcng explans only a fracton of the observed ntal return durng hot ssue perods. Francs and Hasan (001) use SFA to compare the underprcng of venture captal backed wth non-venture captal backed IPOs to test exstng evdence n the lterature that suggests the degree of ntal return depends on venture captal (VC) backng. The offer prce correlates wth frm sze, offer proceeds, the rato of underwrter fee to offer sze, underwrter reputaton, equty retenton by pre-ipo owners, and NYSE as the place of flotaton. The authors estmate a combned fronter on a sample of 843 U.S. IPOs between 1990 and 1993, as well as separate fronters for VC backed and non-vc backed new ssues. Estmates of the average delberate underprcng s 9.1% for the pooled sample, whle the fgures for VC backed and non-vc backed IPOs are 35.3% and 7.1% respectvely. A follow up regresson shows a postve and statstcally sgnfcant assocaton between ntal return and delberate underprcng. Dsaggregatng these estmates nto sze decles suggests that delberate underprcng decreases wth ncreased frm sze. Koop and L (001) apply Bayesan SFA to nvestgate the valuaton of a pooled sample of,969 IPOs and 3,771 seasoned equty offerngs (SEOs) n the U.S. between 1985 and 1998. Ther study departs from Hunt-McCool, Koh and Francs' (1996) s approach n two ways. Frstly, Koop and L (001) argue that SFA usng IPOs only s lkely to result n these ssues beng more effcently prced whch leads to underprcng beng underestmated. Therefore, Koop and L (001) recommend constructng the prcng fronter from both IPOs and SEOs. Ths approach s more lkely to place SEOs closer to the fronter, whle IPOs would be further away from the fronter, resultng n hgher levels of IPO underprcng. Koop and L (001) therefore analyse whether IPOs are underprced per se n relaton to SEOs, rather than whether some IPOs are underprced more than other new ssues. Secondly, they use market captalzaton as an aggregate measure of frm value rather than offer prce per share whch depends on the number of shares n free float. Another contrbuton of ther study nvolves the explct separaton of prcng from underprcng factors n estmaton. The 6
prcng factors affect the fronter and nclude net ncome, revenue, earnngs per share, total assets, debt, and underwrter fees. The underprcng factors nclude dummy varables to capture SEOs, hot and cold new ssues perods, NBER peak ssue months, NYSE new ssues, multple offers made by ssuers, and the top fve lead managers actng as underwrters ranked by market share. The results reveal a postve assocaton between frm value, net ncome, revenues, total assets, and underwrter fees, whle the correlaton between frm value and leverage s negatve. Frms operatng n ndustry sectors wth hgher earnngs potental such as, for example, chemcal products, computers, electronc equpment, scentfc nstruments and communcatons also have hgher valuatons. Among the underprcng factors, only the dummy varables on SEO and multple offers made are postve and statstcally sgnfcant. IPOs are, on average, found to be undervalued by between 5% and 30%, whle SEOs, whch drve the fronter, are undervalued by at most 5%. Chen, Hung and Wu (00) apply SFA to measure delberate premarket underprcng and aftermarket msprcng for a sample of 196 Tawanese IPOs spannng the perod 1985 to 1995. These authors use as prcng factors frm age, offer proceeds, equty retaned by pre-ipo owners, earnngs per share, total assets, and the ndustry prce-earnngs rato. The study reports an average delberate premarket underprcng of 7.46% for the pooled sample, whle the fgures for hot and cold new ssues perods are 7.16 and 7.7%, respectvely. The authors fal to fnd any sgnfcant relatonshp n a regresson of adjusted ntal return 7 on delberate underprcng. In a further analyss, they nvestgate whether nosy tradng n the aftermarket explans ntal return. Nosy tradng s proxed by varables measurng the probablty of recevng an allocaton of IPO shares, the proporton of shares retaned by pre-ipo owners, and frm sze. All three varables have a postve and sgnfcant assocaton wth the 30-day adjusted ntal return. In a further test, the authors nvestgate whether IPOs suffer from poor returns n the long term after nosy tradng actvtes subsde but could not fnd evdence of a postve correlaton n a regresson of 3 year holdng returns on varables proxyng nosy tradng. Chan, Wu and Kwok (007) use SFA to study the mpact of global offer prces on the prcng of U.S. IPOs usng a sample of,307 new ssues between 1986 and 1998. The study uses both prcng factors and underprcng factors to estmate the effcent fronter of offer prces. Prcng factors comprse earnngs per share, the standard devaton of the frm's frst 10 daly returns, offer prce revson, gross spread as a percentage of offer sze, the rank of the lead underwrter, a spn-off dummy varable, dummy varables ndcatng the stock exchange (NYSE, AMEX, NASDAQ), market captalzaton, and ndustry sector dummy varables. The underprcng factor s a dummy varable capturng global IPOs. The estmaton follows the Batesse and Coell (1995) condtonal mean 7 These are cumulatve seven day (as well as thrty day and three year) returns less the estmated delberate premarket underprcng component. 7
model, revealng that purely domestc IPOs are underprced by 3.1 percentage ponts more than global IPOs. Peng and Wang (007) employ SFA and focus on the mpact of offer methods on offer prces usng a sample of 647 Tawanese IPOs between 1996 and 003. They use a two-stage approach to estmatng underprcng of IPOs and ts determnants. In the frst stage, they estmate a stochastc fronter usng earnngs per share, total assets, frm age, debt rato, and proporton of stocks retaned by pre-ipo owners at flotaton as prcng factors. The frst stage s estmated by the method of moment approach. In the second stage, the authors run a regresson of the estmates of underprcng (from the frst stage) on a number of determnants. The underprcng factors nclude offer methods, return and volatlty of market ndex, probablty of stock allocaton n oversubscrbed offerngs, offer sze, amount of sales, level of earnngs management, underwrter reputaton, audtor reputaton, ndustry dummy varables, and stock market dummy varables. A key fndng s that the aucton method reduces the average underprcng when compared to fxedprce offerngs. In addton, the study reports that greater earnngs potental, less rsk or offers wth less nformaton asymmetry have lower underprcng. Fnally, n a recent paper, Luo and Ouyang (014) use Bayesan SFA to analyse the prcng effcency of 379 IPOs n the ChNext market 8 between 009 and 014. They base ther nput varables on fundamentals of ssuers, nformaton asymmetry surroundng IPO value, offer characterstcs, and market envronment. The authors clam that underwrters are predomnantly responsble for prcng neffcency n the ChNext market. Although all the above studes use SFA to measure delberate premarket underprcng, t s noteworthy that, wth the excepton of Koop and L (001), there s no study that explctly separates out prcng factors and controls for underprcng factors n the IPO prcng fronter. In addton, none of the exstng studes attempts to relate the delberate premarket underprcng to proxy varables that capture nformaton asymmetry between market partcpants, whch s one of the domnant explanatons of persstent average ntal return. Exstng studes whch use such proxy varables use them as part of the prcng factors, nstead of treatng these proxy varables as affectng the degree of delberate underprcng. Our paper thus makes several contrbutons to the exstng lterature. We separate out prcng factors and underprcng factors and control for these n the IPO prcng fronter. The effcent prcng fronter enables us to determne the degree of delberate premarket underprcng. We nclude commonly used proxy varables that reflect nformaton asymmetry between market partcpants to explan varatons n delberate premarket underprcng. Dsentanglng the 8 ChNext was launched n October 009 n Chna to provde a new fnancng platform to rase captal for small- and medum-szed enterprses. 8
delberate premarket underprcng from the observed ntal return enables us to also examne whether tradng actvty n the early aftermarket can explan the resdual msprcng component. 3. Offer prce, prcng factors and underprcng factors In lne wth most of the earler studes quoted above, the offer prce per share s the dependent varable n our stochastc fronter. The ndependent varables are prcng factors, whch are prmary drvers of equty value and orgnate from standard fnance theory. Estmates of neffcency reflect delberate premarket underprcng and varatons n these are related to proxy varables that capture varous types of nformaton asymmetres surroundng IPO value between partcpants n the new ssues market. 3.1. Prcng factors Estmatng expected cash flow, cost of captal and growth opportuntes to value the equty of IPOs s nontrval. Relatvely lttle s known about the past performance of IPO frms and ther future prospects. Instead of attemptng to estmate expected cash flow, cost of captal and growth opportuntes drectly, we use several proxy measures. We use earnngs before nterest, tax, deprecaton and amortzaton (EBITDA) n the most recent twelve-month perod before gong publc as an nput nto our valuaton model even though fnance theory advocates usng cash flows. EBITDA s a better predctor of future operatng cash flows than current operatng cash flows as evdenced n prevous research. Teoh, Welch and Wong (1998) fnd that earnngs better reflect a more long-term stream of wealth creaton than cash flows that are subject to hgher annual volatlty. Also, IPO prospectuses do not normally dsclose objectve (audted) cash flow forecasts and estmates of cost of captal. In the long-term, earnngs converge to cash flows. We rely on EBITDA as opposed to earnngs per share (EPS) to crcumvent clams that IPO frms rase reported EPS by manpulatng dscretonary accountng accruals before gong publc. Teoh, Welch and Wong (1998) fnd that excessve earnngs management can lead to hgher ntal valuatons, followed by long-run underperformance. Indeed, Koop and L (001) confrm that EPS s not an mportant prcng factor n IPO valuaton. In a more recent study, Aggarwal, Bhagat and Rangan (009) therefore use EBITDA n ther valuaton model as opposed to earler studes that rely on EPS (Koop and L, 001; Chen, Heng and Wu, 00; Peng and Wang, 007; Luo and Ouyang, 014). Industry sector dummy varables take account of dfferences n the cost of captal and busness rsk across ndustry sectors. We classfy the frms nto 1 ndustry sectors usng the two-dgt SIC level, n lne wth Koop and L (001). In the absence of any stock market tradng n the perod leadng up to flotaton and any dsclosure of objectve (audted) estmates of the cost of captal n 9
the flotaton prospectus, we use these ndustry sector dummy varables to reflect the dvergent busness rsk. Three sets of proxy measures capture earnngs potental and growth opportuntes that are key contrbutors to IPO value. The frst proxy measure s a dummy varable that takes account of negatve earnngs n the accountng perod before gong publc. Aggarwal, Bhagat and Rangan (009) argue that negatve earnngs represent valuable future growth opportuntes rather than current proftablty. In ther study, frms wth hgher negatve earnngs have hgher valuatons, whch would appear to be counter-ntutve at frst glance from a proftablty pont of vew. The amount of R&D spendng n the accountng perod before gong publc s the second proxy measure. Aggarwal, Bhagat and Rangan (009), Hertzel, Huson and Parrno (01) and Deeds, Decarols and Coombs, (1997) use R&D nvestment to proxy for growth opportuntes and hence value drver. Our ndustry sector dummy varables dscussed above are the thrd proxy measure. They take account for dfferences n earnngs potental and growth opportuntes across varous ndustres. Hunt- McCool, Koh and Francs (1996), Koop and L (001) also control for ndustry sector afflaton n ther stochastc fronter model. We use book value of assets n the most recent twelve-month accountng perod before gong publc as the lower bound for IPO value. Hunt-McCool, Koh and Francs (1996), Koop and L (001), Chen, Hung and Wu (00), and Peng and Wang (007) also nclude the book value of assets as a prcng factor n ther value estmatons. Fnally, we use fnancal leverage, measured by long-term debt dvded by book value of assets n the accountng perod before gong publc as proxy for rsk. Frms wth hgher fnancal dstress have a greater chance of gong bankrupt and hence should have lower valuatons. Koop and L (001) and Peng and Wang (007) corroborate a negatve relatonshp between leverage and frm value. 3.. Underprcng factors Exstng research explans persstent average ntal return as the outcome of nformaton asymmetry on IPO value between ssuers, underwrters and dfferent nvestor segments. Our underprcng factors reflect these nformaton asymmetres at varous stages of the IPO valuaton process. We nclude sales and age of the ssung frm to control for ex ante uncertanty surroundng IPO value. These varables measure how establshed a frm s. Smaller frms wth lower growth n sales and shorter operatng hstory suffer from greater ex ante uncertanty surroundng IPO value. Greater ex ante uncertanty forces ssuers to underprce IPOs to compensate unnformed nvestors for an adverse selecton problem. Unnformed nvestors end up wth a hgher proporton of 10
overprced ssues than nformed nvestors who only apply for underprced IPOs (Rock, 1986; Beatty and Rtter, 1986). Accordngly, we use the amount of sales n the most recent twelve-month accountng perod before gong publc. Hunt-McCool, Koh and Francs (1996), Koop and L (001), Peng and Wang (007), and Aggarwal, Bhagat and Rangan (009) report a postve assocaton between sales and IPO value. Frm age, measured n years between the frm's date of foundaton and the IPO date, has a postve assocaton wth IPO value accordng to the fndngs of Hunt- McCool, Koh and Francs (1996). We use equty retaned by pre-ipo shareholders after the ssue, proceeds rased at the dsposal of the ssung frm, the commtment of pre-ipo shareholders not to sell addtonal shares mmedately n the aftermath of the ssue, and underwrtng fees as proxy measures for ex ante uncertanty surroundng the deal characterstcs. It s not unreasonable to assume that frm nsders know more about future frm prospects than outsde nvestors (Akerlof, 1970). Frm nsders can transmt costly and dffcult to mtate sgnals to convey frm value to outsde nvestors (Spence, 1973). We use Leland and Pyle s (1977) equty retaned by pre-ipo shareholders n the post-ipo frm jontly wth proceeds at the dsposal of the ssuer as sgnal n whch frm nsders transmt nformaton on frm value to outsde nvestors. The mpact of these sgnallng varables on the stochastc fronter prce s nconclusve. Some studes fnd a relatonshp whereas other studes report nsgnfcant results or omt these sgnallng varables altogether. Our study therefore seeks to re-examne the mportance of these sgnallng varables. The proporton of equty retaned s a costly sgnal because owners forgo the possblty of achevng dversfcaton of ther personal nvestment portfolo. Varous conceptual frameworks use equty retaned as sgnal to reduce nformaton asymmetry surroundng frm value (Grnblatt and Hwang, 1989; Welch, 1989; Allen and Faulhaber, 1989). Hunt-McCool, Koh and Francs (1996) report a postve relatonshp between equty retaned and estmated offer prce. Van der Goot and Roosenboom (005) show a postve assocaton between management stock ownershp and market captalzaton. Aggarwal, Bhagat and Rangan (009) fnd a negatve relatonshp between nsder retenton and offer value. Chen, Hung and Wu (00) do not fnd a statstcally sgnfcant relatonshp. Koop and L (001) and Luo and Ouyang (014) omt equty retenton as a sgnallng varable n ther stochastc fronter model. The amount of IPO proceeds at the dsposal of the ssung frm s an effectve sgnal, perhaps of nvestment plans, for example. Studes normally use proceeds at the dsposal of the ssung frm n conjuncton wth equty retaned by pre-ipo owners as a jont sgnal of frm value (Downes and Henkel, 198; Rtter 1984b; Km and Wesbach, 008). Hunt-McCool, Koh and Francs (1996) report a postve relatonshp between proceeds and offer prce, whle Peng and Wang (007) report a negatve assocaton between the amount of money rased and prcng neffcency. Koop and L (001), Chen, Hung and Wu (00) and Aggarwal, Bhagat and Rangan (009) omt proceeds as a sgnallng varable n ther stochastc fronter models. 11
We use a dummy varable to account for the presence of lock-n agreements. The mpact of lockn agreements on prces n a stochastc fronter framework awats nvestgaton. A lock-n agreement s a commtment that prohbts frm nsders from sellng shares n the aftermarket for a specfed perod. Demand n IPO shares wth a lock-n agreement should be hgher than offers wthout such an agreement. Investors are more wllng to buy an IPO knowng that they have a reduced moral hazard problem durng the perod n whch frm nsders cannot sell ther equty stakes. Ths argument s n lne wth the fndngs of Brav and Gompers (003) and Arthurs, Busentz, Hosksson and Johnson (009). The supply of tradable shares ncreases only followng the lock-n expraton, whch could potentally have a negatve mpact on share prces (Bradley, Jordan, Ha-Chn and Roten, 001; Feld and Hanka, 001). In the US, the lock-n perod can be as short as 90 days and as long as two years. We use underwrtng expenses as a proxy for the rskness of an IPO. Underwrter compensaton s lkely to be a functon of nformaton costs and deal characterstcs (Hughes, 1986). Fees wll be hgher for those IPOs about whch relatvely lttle publc nformaton s avalable such as, for example, less well-establshed and younger frms. Accordngly, varous studes use underwrter expenses as proxy for ex ante uncertanty surroundng the deal characterstcs. In a stochastc fronter framework, Hunt-McCool, Koh and Francs (1996) report a negatve assocaton between underwrter compensaton and estmated offer prce. In contrast, Koop and L (001) report a postve relatonshp between fees and market captalzaton at the offer prce. Nether of the two studes smultaneously control for underwrtng fees and underwrter reputaton. Chen, Hung and Wu (00) and Aggarwal, Bhagat and Rangan (009) do not take account of underwrtng expenses n ther analyss. We therefore provde further evdence on the smultaneous effect of underwrter fees and reputaton to shed lght on earler mxed fndngs. We use logarthm of one plus the number of uses of IPO proceeds dsclosed n the flotaton prospectus as proxy of ex ante uncertanty surroundng frm value. Ths measure s to some extent endogenous and hence we smultaneously control for deal characterstcs. The number of uses of proceeds s a functon of offer sze, the proporton of the proceeds at the dsposal of the frm and the proceeds gong to sellng pre-ipo shareholders. Therefore, a smple count of the number of uses of proceeds wthout controllng for offer sze and the proporton of funds at the dsposal of the ssuer s unlkely to be nformatve and can lead to contradctory fndngs n a cross-sectonal analyss. For example, Beatty and Rtter (1986) report a postve assocaton between the number of uses of proceeds and underprcng. In contrast, Ljungqvst and Wlhelm (003) and Leone, Rock and Wllenborg (007) fnd that more specfc dsclosure leads to lower underprcng. The absence of the number of uses of proceeds from studes usng SFA to explan delberate underprcng allows us to extend the lterature. 1
We use underwrter reputaton, prvate-equty and/or venture captal backng as thrd-party certfcaton of ex ante uncertanty to endorse IPO value. Underwrter reputaton employs Loughran and Rtter s (004) rankng of nvestment banks n tombstone advertsements of IPO prospectuses. Regular updates on underwrter reputaton are avalable from Jay Rtter s web ste. These updates allow for varaton n underwrter reputaton across dfferent nvestment banks and tme perods. Reputaton captal and the correspondng rankng can mprove, deterorate or reman unchanged. We beleve that tombstone rankngs are perhaps the most meanngful ndcators of reputaton captal. Frstly, the rankngs n tombstone advertsements most closely reflect underwrter actvty n the IPO market. Underwrter rankngs have not prevously been used n SFA applcatons to explan delberate premarket underprcng. Secondly, market share (see, for example, Meggnson and Wess, 1991; Beatty, Bunss and Hand, 1998; Habb and Ljungqvst, 001; Chemmanur and Paegls, 005), s a more ambguous defnton of reputaton snce t nvolves a trade-off between usng aggregate number of IPOs or aggregate offer sze as the bass for any rankngs. Thrdly, underwrter fees (see, for example Booth and Smth, 1986; Hunt-McCool, Koh and Francs, 1996; Koop and L, 001) may not be a representatve proxy of reputaton captal gven the sample perod under nvestgaton. Hsuan-Ch and Rtter (000) provde evdence that underwrtng fees are clustered around 7% for frm-commtment offers durng the late 1990s, whch overlaps wth our sample perod. Any devaton from 7% s therefore lkely to be assocated wth dfferences n the rsk of a deal rather than a reflecton of the reputaton captal of underwrters. We employ a dummy varable to capture IPOs wth prvate equty backng and/or venture captal backng. Our study s the frst analyss to test the mpact of prvate equty and/or venture captal backng on the delberate premarket underprcng n a stochastc fronter context. Dsentanglng the relatve mpact of prvate equty backng and/or venture captal backng on delberate underprcng s mportant because of the mxed results reported n the lterature explanng ntal return. On the one hand, some studes clam that prvate equty backng and/or venture captal backng endorses frm value and reduces ntal return (see, for example, Barry, Muscarella, Peavy and Vetsuypens, 1990; Meggnson and Wess, 1991; Lerner, 1994; Brav and Gompers, 1997; Nahata, 008; Arthurs, Busentz, Hosksson and Johnson, 009; Bartlng and Park, 009; Krshnan, Ivanov, Masuls and Sngh, 011; Nanda and Rhodes-Kropf, 013). Furthermore, younger venture captal backed IPOs have, on average, hgher ntal return than those IPOs that have more establshed venture captal backng (Gompers, 1996). On the other hand, Lu and Rtter (011) report that venture captal backed IPOs have hgher ntal return when they have coverage from more reputable analysts. We use aggregate IPO number and aggregate average ntal return to take account of new ssues market condtons at the tme of flotaton. Data on aggregate IPO number and aggregate average ntal return are avalable from Jay Rtter s web ste. We follow Yung, Çolak and We s (008) 13
approach to dscrmnate between hot, cold and normal new ssues market actvty. Hot market actvty s when the number of IPOs n a quarter s 50% greater than the three-monthly movng average. Cold market actvty s when the number of IPOs n a quarter s 50% smaller than the three-monthly movng average. All other quarters fall nto the category of normal new ssues market actvty. Our analyss s the frst study to nclude new ssues market condtons n a stochastc fronter framework to test the mpact on delberate underprcng. We nclude aggregate IPO number and aggregate average ntal return because Lowry (004), Loughran and Rtter (004), Bralsford, Heaney and Sh (009), and Yung, Çolak and We (008) dentfy two stylsed facts about these measures. Frstly, the number of IPOs and average ntal return are hghly autocorrelated. Secondly, the correlaton between the two seres s postve. Fnally, we use the change n real prvate nonresdental fxed nvestment as proxy for prvate frms demand for captal. We use the same approach as above to dscrmnate between hot, cold and normal perods for demand for captal. The mpact of ths varable on delberate underprcng n a stochastc fronter analyss awats nvestgaton. Yung, Çolak and We (008) measure nvestment opportuntes drectly va the change n real prvate nonresdental fxed nvestment rather than ndrectly va fnancal statements at frm level to nfer the value of nvestment opportuntes. Frm attrbutes or deal characterstcs do not suffcently dffer to fully account for the IPO cycles (Rtter, 1984a; Lowry and Schwert, 00; Loughran and Rtter, 004). Yung, Çolak and We (008) demonstrate that exogenous shocks to the value of prvate frms nvestment opportuntes can lead to more actve IPO markets wth more low qualty ssuers gong publc than the average pre-shock ssung frm. These exogenous shocks can therefore lead to waves wth heghtened dsperson n frm qualty, hgher nformaton asymmetry surroundng frm value and hence hgher average ntal return durng hot IPO markets. 4. Early aftermarket tradng In addton to nformaton asymmetry surroundng IPO value between market partcpants n the prmary market, the lterature attrbutes persstent average ntal return also to early aftermarket tradng actvty (see, for example, Chen, Hung and Wu, 00; Ells 006; Agarwal, 003). Persstent delberate underprcng n the prmary market creates excess demand n the aftermarket when IPO shares become avalable for tradng. Ths excess demand wll push up prces to a new equlbrum and allows sellng shareholders n the secondary market to earn sgnfcant returns. We refer to the dfference between the market prce and the 'far' offer prce as msprcng n the aftermarket. Fve varables explan msprcng n the early aftermarket. The number of shares traded on the frst day as percentage of shares offered serves as a proxy varable to take account of the tradng volume and excess demand of IPOs n the aftermarket. Exstng research reports of a postve assocaton between ntal return and tradng volume n the 14
aftermarket (Aggarwal, 000, 003; Chahne, 007; Ells, Mchaely and O Hara, 000, 00; Ells, 006). Two dummy varables capture the aftermarket demand n IPO shares va the prce revson from the ntal flng range to the offer prce (Hanley, 1993) as an ndcator of the lkelhood of flppng takng place. We code dummy varables equal to one f the offer prce s above or below the ntal flng prce range. More nvestors buy or sell shares f shares are offered above or below the ntal flng prce range. Ells (006) observes a sgnfcant assocaton between these two dummy varables and tradng volume. Equty retenton by pre-ipo owners mmedately after flotaton serves as a proxy measure for ownershp concentraton. Average ntal return s attrbutable to a sgnallng effect by frm nsders to outsde nvestors. Chen, Hung and Wu (00) report a negatve assocaton between ownershp concentraton and aftermarket msprcng. Underwrter reputaton captures dstrbuton networks and nsttutonal contacts. Ells (006) reports greater tradng volumes for more reputable underwrters. The author argues that hgher qualty underwrters have a greater ablty to generate nterest among nvestors through more establshed dstrbuton networks and nsttutonal contacts. In addton, more reputable underwrters have a hgher reputaton captal at stake and hence are more lkely to provde prce support n the aftermarket. Offer sze captures the rsk of an IPO. Investors perceve smaller IPOs as rsker compared to larger new ssues of more establshed companes. Beatty and Rtter (1986), Rtter (1987), and Carter (199) observe a negatve assocaton between offer sze and ntal return. 5. Data and methods 5.1. Data Our ntal source s the New Issues database n Thomson One Banker, from whch we obtan 3,131 observatons wth all relevant varables on U.S. IPOs from January 1980 to December 01. Ths sample excludes closed-end funds, Amercan Depostory Recepts (ADRs), Master Lmted Partnershps (MLPs), Real Estate Investment Trusts, unt offers, all new ssues wth an offer prce below $5, and IPOs wth an offer prce greater than $1000. Data tems we obtan from Thomson One nclude ssung frm attrbutes, deal characterstcs, and aftermarket tradng data. We use fnancal statement data from Compustat and tradng data for each IPO frm from the Centre for Research n Securty Prces (CRSP) to complement our varables lst and fll n mssng prce observatons n Thomson One. 9 Jay Rtter's web ste provdes the rankngs on underwrter 9 Key fnancal statement data leadng up to flotaton was mssng for many frms n Thomson One and Compustat. 15
1980q1 1981q1 198q1 1983q1 1984q1 1985q1 1986q1 1987q1 1988q1 1989q1 1990q1 1991q1 199q1 1993q1 1994q1 1995q1 1996q1 1997q1 1998q1 1999q1 000q1 001q1 00q1 003q1 004q1 005q1 006q1 007q1 008q1 009q1 010q1 011q1 01q1 013q1 Average Intal Returns (%) reputaton and aggregate data on the new equty ssues market. The Federal Reserve Bank of St. Lous s our last data source whch supples data on prvate non-resdental fxed nvestment across the sample perod to dentfy frms' demand for prvate captal. Our sample excludes observatons wth mssng values on the varables of nterest. Fgure graphs number of IPOs aganst average ntal return for our sample. Overall, the cyclcty s representatve of the data reported on Jay Rtter's web ste. Fgure : Average ntal return and number of U.S. IPOs by quarterly ntervals, 198001 Average ntal return per quarter s the equally-weghted mean of the percentage change from the offer prce n the premarket to the aftermarket prce. 100 80 60 40 0 0 110 100 90 80 70 60 50 40 30 0 10 0-10 -0 Year-Quarter Number of IPOs Average Intal Returns (%) Table 1 shows the sample dstrbuton of IPOs across our ndustry sector classfcaton at the twodgt SIC level. Both the number of IPOs and average ntal return dffer across ndustry sectors. There s no perfect correlaton between the number of IPOs and average ntal return. 'Computers' has the hghest concentraton n terms of the number of offers as well as the hghest average ntal return. 'Utltes' has the lowest concentraton of new ssues, whle 'Ol and Gas' record the lowest average ntal return across the sectors. 16
Table 1: IPO sample dstrbuton across ndustry sectors Our sample conssts of 3,131 offers between 1980 and 01, classfed nto Standard Industral Classfcaton (SIC) sectors correspondng to Koop and L s (001) ndustry classfcaton. Industry Two-dgt SIC codes Number of IPOs Average Intal return (%) Chemcal Product 8 199 9.436 Communcatons 48 115 16.050 Computers 35,73 754 34.939 Electronc Equpment 36 39 7.858 Fnancal Servces 60-65, 67 38 14.844 Health 80 89 9.466 Manufacturng 30-34 115 15.945 Ol and Gas 13,9 97 5.500 Retal 53,54,56,57,59 09 11.053 Scentfc Instruments 38 199 16.345 Transportaton 37, 39, 40-4, 44, 45 157 11.978 Utltes 49 56 9.188 All Others 1,, 7, 10, 1, 14, 664 1.87 Total 3131 19.085 5.. Defnton and measurement of varables Table lsts the varables, ther defntons and measurements as well as the correspondng data sources. Offer prce per share s the prmary varable of nterest We employ several prcng factors as the prmary value drver for equty. Earnngs before nterest, tax, deprecaton and amortzaton s the proxy measure for a more long-term stream of operatng cash flow because the latter s subject to hgher annual volatlty. Industry dummy varables reflect Koop and L's (001) 1 ndustry sectors based on Fama and French's (1997) two-dgt classfcaton. These ndustry sector dummy varables take account of dfferences n the cost of captal, busness rsk, future earnngs potental, and growth opportuntes of IPO frms across ndustres. A dummy varable captures negatve earnngs n the accountng perod before gong publc as a proxy for future growth opportuntes rather than current proftablty. We also nclude hstorc R&D spendng as an ndcator for future growth opportuntes. Book value of assets n the most recent twelve-month accountng perod before gong publc acts as the lower bound for IPO value. Fnancal leverage s the last prcng factor and measures fnancal dstress of frms. We use varous factors to explan varatons n delberate premarket underprcng. Issung frm attrbutes that take account of ex ante uncertanty nclude sales n the accountng perod before flotaton and frm age. In addton to frm attrbutes, we nclude several deal characterstcs as proxy measures to explan delberate premarket underprcng. They nclude equty retaned by pre-ipo shareholders after the ssue, proceeds rased at the dsposal of the ssung frm, the 17
commtment of pre-ipo owners not to sell addtonal shares mmedately n the aftermath of the ssue, underwrtng fees, and logarthm of one plus the number of uses of proceeds dsclosed n the Table : Varable defntons, measurements and data sources Ths table presents the defntons of the dependent and ndependent varables used n the SFA model and the later model for aftermarket msprcng. For the SFA model, offer prce per share s the varable of prmary nterest. Prcng factors and underprcng factors are the ndependent varables. The prcng factors represent the prmary value drvers of equty. The underprcng factors explan the delberate premarket underprcng and these nclude proxy varables relatng to ssung frm attrbutes, deal (offer) characterstcs, thrd-party certfcaton, new ssues market condtons, and prvate frms' demand for captal. For the aftermarket msprcng model, the dependent varable s aftermarket msprcng retreved after estmaton of the SFA model. Ths aftermarket msprcng component s explaned by a number of varables proxyng tradng volume, lkelhood of flppng, ownershp concentraton, underwrter reputaton and offer rsk. Data sources nclude Thomson One Banker, Compustat, Centre for Research n Securty Prces (CRSP), Jay Rtter's web ste [http://bear.warrngton.ufl.edu/rtter/podata.htm], Fama and French's (1997) ndustry sector classfcaton, and the Federal Reserve Bank of St. Lous. Varables used n the Stochastc Fronter Model Dependent Varable Source Offer prce Offer prce per share n US$. Thomson EBITDA Industry dummes Negatve net ncome Captal expendture Prcng Factors Earnngs before nterest, tax, deprecaton and amortzaton (EBITDA) n US$ mllon n the accountng perod before IPO. Industry sector dummy varables as proxy varables for dfferences n the cost of captal and earnngs potental between ndustry sectors, based on Fama and French s (1997) ndustry classfcaton at the two-dgt level to dscrmnate between IPOs comng from 1 dfferent ndustry sectors n lne wth Koop and L (001). Dummy varable coded one f the frm has a negatve net ncome n the accountng perod before gong publc, else coded zero. R&D nvestment n US$ mllon n the accountng perod before IPO. Compustat Fama and French (1997), Koop and L (001) Compustat Compustat Assets Total assets n US$ mllon n the accountng perod before IPO. Thomson, Compustat Leverage Issung frm attrbutes Long-term debt scaled by book values of total assets n the accountng perod before IPO. Underprcng Factors Compustat Sales Sales n US$ mllon n the accountng perod before IPO. Compustat Frm Age Number of years between the date the company was founded and the IPO date. Compustat Deal (offer) characterstcs Equty retaned Proceeds Lock-n Leland and Pyle s (1977) sgnal of equty retaned:, where RET Thomson s the proporton of equty retaned by pre-ipo shareholders n the post-ipo frm. Proceeds n US$ mllon at the dsposal of the ssung frm, measured by the number of prmary shares tmes the offer prce per share n US$. Dummy varable coded one f the flotaton prospectus dscloses the presence of a lockn agreement, else coded zero. Thomson Thomson Fees Underwrtng fees n US$ mllon. Thomson ln(1 + number of uses of proceeds) Thrd-party certfcaton Reputaton Logarthm of one plus the number of uses of IPO proceeds dsclosed n the flotaton prospectus. Underwrter reputaton based on tombstone rankngs used n Carter and Manaster (1990), Carter and Dark (1993) and Carter, Dark and Sngh (1998). Updated rankngs are avalable from Jay Rtter s web page. Thomson Jay Rtter s IPO webste 18
VC/PE backng IPO market condtons Hot or cold market actvty Hot or cold market underprcng Dummy varable coded one f pre-ipo venture captalsts and/or prvate equty nvestors retan a stake n the post-ipo frm, else coded zero. Two dummy varables capture IPO market actvty durng hot or cold markets. Hot market actvty s coded one f the number of IPOs n a quarter s 50% greater than the three-monthly movng average. Cold market actvty s coded one f the number of IPOs n a quarter s 50% smaller than the three-monthly movng average. All other quarters fall nto the normal market actvty and coded as zero. Two dummy varables capture IPO market underprcng durng hot or cold markets. Hot market underprcng s coded one f the average IPO underprcng n a quarter s 50% greater than the three-monthly movng average. Cold market underprcng s coded one f the average IPO underprcng n a quarter s 50% smaller than the threemonthly movng average. All other quarters fall nto the normal underprcng category and coded zero. Prvate frms demand for captal Hot or cold demand for captal Aftermarket Msprcng Tradng volume % of shares traded Two dummy varables capture prvate frms demand for captal durng hot or cold markets. Hot demand for captal s coded one f the quarterly percentage change n real prvate nonresdental fxed nvestment s 50% greater than the three-monthly movng average. Cold demand for captal s coded one f the quarterly percentage change n real prvate nonresdental fxed nvestment s 50% smaller than the three-monthly movng average. Aftermarket msprcng component of delberate underprcng Varables used n Aftermarket Msprcng Model Dependent Varable ntal returns recovered after estmaton of Independent Varable The number of shares traded on the IPO day dvded by the total number of shares n the offer. Thomson Jay Rtter s IPO webste Jay Rtter s IPO webste Federal Reserve Bank of St. Lous Authors calculatons CRSP, Thomson Lkelhood of flppng Above flng range Dummy varable equal to one f the offer prce s above the ntal flng prce range, else coded zero. Thomson Below flng range Ownershp concentraton Equty retaned Underwrter Reputaton Reputaton Offer rsk Dummy varable equal to one f the offer prce s below the ntal flng prce range, else coded zero. Thomson Leland and Pyle s (1977) sgnal of equty retaned:, where RET Thomson s the proporton of equty retaned by pre-ipo shareholders n the post-ipo frm. Underwrter reputaton based on tombstone rankngs used n Carter and Manaster (1990), Carter and Dark (1993) and Carter, Dark and Sngh (1998). Updated rankngs are avalable from Jay Rtter s web page. Offer sze Total number of shares offered n the IPO multpled by the offer prce. Thomson Thomson, Jay Rtter's IPO webste flotaton prospectus. Furthermore, we use thrd-party certfcaton to explan delberate premarket underprcng. Underwrter reputaton rankngs are avalable from Jay Rtter's web ste. We use the average rank of underwrters for those IPO deals that have more than one underwrter. A dummy varable captures IPOs wth prvate equty backng and/or venture captal backng. To account for IPO market condtons, we use Yung, Çolak and We's (008) approach to classfy aggregate IPO 19