2487752

7/30/2019 2487752

1/10

A Tax-Based Motive for the Underpricing of Initial Public OfferingsAuthor(s): Mary Ann Reside, Richard M. Robinson, Arun J. Prakash, Krishnan DandapaniReviewed work(s):Source: Managerial and Decision Economics, Vol. 15, No. 6 (Nov. - Dec., 1994), pp. 553-561Published by: John Wiley & SonsStable URL: http://www.jstor.org/stable/2487752 .

Accessed: 06/03/2012 04:27

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

John Wiley & Sons is collaborating with JSTOR to digitize, preserve and extend access toManagerial and

Decision Economics.

http://www.jstor.org

7/30/2019 2487752

2/10

MANAGERLALAND DECISION ECONOMICS, VOL. 15, 553-561 (1994)

A Tax-Based M o t i v e F o r T h e

Underpricing O f I n i t i a lP u b l i c

OfferingsMaryAnn Reside and RichardM. Robinson

EasternKentucky University,Richmond,KY USAArun J. Prakashand KrishnanDandapani

Florida International University,Miami, FL, USAThis paper presents a modelof entrepreneurialwealth maximizationor the pricingofinitial public offerings (IPOs). It is an extensionof one previouslypresentedin theliterature.The model showsthat personal ax rateson ordinaryncomeandcapital gainsmay, in part, determineIPO pricing: an increase in the capital gains tax rate shouldlower the degreeof underpricing.An empiricalanalysisof the effectof the Tax ReformAct of 1986, which raised the capital gains tax rate, shows that the averagedegree ofunderpricingdid decrease as predicted,and that this occurs aftercontrolling or otherpossible influences.

INTRODUCTIONNumerous studies have documented significantunderpricing f IPOs. In a recent article in thisjournal, Dandapani et al. (1992, hereinafterDDRP) provide yet another explanation,basedon personal taxes, for the underpricing f initalpublicofferings IPOs).The 'traditional' ypothe-ses for the underpricingphenomenon includerisk-averse underwriter/risk compensation,monopsonypower, agencycost, institutional ag,speculativebubble, implicitinsurance and asym-metric information.See DDPR for a review ofthese hypotheses.Addingto these traditionalhypotheses,DDPRproposea model by whichunderpricings depen-dent on the entrepreneur's1) personal tax rateon ordinary ncome (and, concomitantly, he ap-plicable capital gains tax rate), (2) abilityto defercapital gains, and (3) the proportionof retainedownership n the firm.Theypresenta theoreticalargument hatthere is a tax-basedmotivefor IPOissuers to underprice.Thispaper extendsthe theoreticalargument ora tax-basedmotive for IPO underpricing.TheDDPR entrepreneurialwealth-maximizingmodel

of IPO pricing is reviewed and a considerableextensionof the model is presented.WhereastheDDPR modelshows the influenceof tax rates onthe proportionof IPOs underpriced, he exten-sion to the model presented here illustratestheinfluenceof tax rateson the degree of underpric-ing of each IPO issue. In addition, an expandedempirical analysisis presented in this paper ascompared o DDPR. This new empiricalanalysisinvolvesa much-expanded ata set, andalso moreextensive parametricand nonparametric nalysisof the tax effects on IPO underpricing.

THEORETICALFOUNDATIONThe Tax-based PO Model of DDPR

The DDPR tax-based model assumes that thepurpose of the IPO is to fund a project withpositivenet presentvalue. To simplify he analy-sis, agency problemsbetweenthe issuersand theunderwriters, and informational asymmetriesbetween the issuersand the market,or betweengroups of investors, are assumed nonexistent.Furthermore,the entrepreneuris assumed re-sponsible or settingthe issue priceof the shares.

CCC 0143-6570/94/060553-09? 1994 by John Wiley & Sons,Ltd.

7/30/2019 2487752

3/10

554 M. A. RESIDE, R. M. ROBINSON, A. J. PRAKASH AND K. DANDAPANIWhilecorrectpricingwould allowfull realizationof the project'snet present value, it may alsoforce the net present value to be received as acurrentlyrealized taxable gain. Underpricing heissue, whichcould result in a currentlyunrealizedcapital gain, may therefore be in the best inter-ests of the entrepreneursince the tax would bedeferred until the gain is realized. The en-trepreneur aces the problem of pricing the IPOso that he or she receives the maximumcurrentwealthfrom the new project.The mntrepreneur ay appropriate he wealthgenerated by the project by paying himself adividend(equal to the net present value of theproject)prior o the issue, while borrowing gainstthe future value of the project n order to financethe dividend.Alternatively,he entrepreneur anchargethe firm a royaltyequal to the net presentvalue and payable rom the proceedsof the issue.Assumingthat there is agreementbetween themarketand the issuers on the value of the issue,then post-issue,the marketwill price the firm atits presentvalueand it willbe indifferent owardsthe method that the entrepreneurmight choosefor appropriatinghis 'surplus'net present value.Each of the above methods, however,whether apriordividendor a royalty chedule, mposes a taxliabilityon the entrepreneur. n addition, or thecase of the entrepreneurwho retains a portionofthe issue, he or she may receive part of hisinvestmentbackas realized taxablegains.The entrepreneur'sdecision on the issue pricewill therefore affecthis or her return n twoways,given that the IPO issue price is set below theproject'spresentvalue, and therefore below themarketvalue. The firstway is that a higherIPOprice increases 'entrepreneurialsurplus' to beappropriated y eithera dividendor royalty.This'surplus' s the differencebetween the project'scost and the IPO price.The secondwayis that ifsome proportionof ownership s retainedby theentrepreneur hroughthe purchaseof shares ofthe IPO at the offer price, then a lower issueprice increases the capital gain to his or hershares.This capital gain occurswhen the shareprice rises in the after-market o reflect the pre-sent value of the fundedproject.In a world with personal taxes, the en-trepreneurial urpluswill be taxed at rates appli-cable to ordinary ncome. Capital gains will betaxed when realized, and at rates applicabletocapital gains income, which prior to 1987 were

typically lower than ordinary ncome tax rates.The theoreticalanalysis hat followsremainsvalid,however, or the current ax laws,wherethe ordi-nary income tax rate equals the capitalgains taxrate, due to deferral capability.The former canlead the entrepreneur o set the issue price belowthe present value of the project so as to receivereturns n the form of capitalgains ratherthan assurplus.The choice of the issue price is modeledbelow. Let:PV= present value of the proposedproject;C = currentperiod cash outflow needed to fi-nance the project;P = total issue price chosen by the entrepreneur

(C < P < PV);a =proportion of firm retained by the en-trepreneur O< a < 1);To = entrepreneur'spersonal tax rate on ordi-nary income;Tk = entrepreneur's ax rate on capitalgains in-come (Tk < To); andy = a present-value nterest factor to measurethe present value of deferredtax on real-ized capital gains (O< y < 1), over the en-trepreneur'sdesireddeferralperiod,where

if y = 1, capital gains cannot be deferredand must be realizedin the currentperiod,and if y= 0, indefinite deferral would bepossible and the capital gains tax could becompletelyavoided.The entrepreneurwishes to maximizehis or herwealth, W,which consists of the entrepreneurialsurplus taxed at To, shown as (P - C) (1 - To) inEqn. (1) below, plus the capital gains taxed atyTk, shown as (PV-P)Ca(1-yTk). The en-

trepreneur olves the maximization roblemgivenby:Max W= (P- C)(1 - To) + (PV-P)a

pX(l-yTk) (1)

This equation is similar to the DDPR linearmodelwhich, given a, resultsin corner solutionsthat indicatewhetheror not underpricing xists,but does not indicatethe degreeof underpricing.The model presented in the next section, how-ever, is nonlinearin that a is a functionof P.Thismodelresultsin solutionsfor the magnitudeof underpricing.

7/30/2019 2487752

4/10

UNDERPRICING OF IPOS 555TheExtendedTax-based PO Model

In orderto ascertainthe degree of underpricing,assume that a, the proportionof entrepreneurialretention, s some unspecified unctionof P. Thatis, allowa to be influencedby the issue price, P,in that the lower the issue price, the higher theproportion f the issue retained.This would occurbecause the lowerthe issue price,the greaterthefuture capital gain to be exploited and the moreattractive he holdingsof the issuewouldbe. Thisrelation s expressedby Eqn. (2), wheref(P) is atleast a twice-differentiable unction of P. Themaximization roblem, herefore,reduces to thatexpressedby Eqn. (3):a=f(P) (2)

where f'(P) < 0:Max W= (P - C)(1 - To) + (PV- P)f(P)

X(1 - yTk) (3)The first-ordercondition for an extremumisgivenby Eqn. (4), and the second-order onditionfor a maximum s given by inequality 5):

dW/dP = (1 - To) -f(P)(1 - yTk)+(PV-P)f'(1 - yTk) = O (4)d2W/dP2 = -2f'(1 - yTk) + (PV-P)f"

X(1-yTk)

7/30/2019 2487752

5/10

556 M. A. RESIDE, R. M. ROBINSON, A. J. PRAKASH AND K. DANDAPANIEquation 11) followsfrom(lOb):

dP{-2f'(1 - yTk) +f"(PV-P)(1 - YTk)}+dy{Tk[ a-(PV-P)f']} = 0 (lOa)

dP(d2W/dP2) + dy{Tk[a - (PV-P)f']} = 0(lOb)dP/dy = Tk[(PV- P)f' - a ]/(d2W/P2)

(11)Since PV> P for underpricing, nd f' < O, > 0and d2W/dP2 < 0 by the second-order onditionfor a maximumt followsthat dP/dy > 0.Assume y = (1 + i)-', where t is the en-trepreneur'sdeferralperiod and i is his or herdiscountrate. Then dy/di = -t(l + -'-1, andsince t > 0 it mustbe that dy/di < 0. Also, sincedP/di = (dP/dyXdy/di), and since dP/dy > 0,anddy/di < 0, then dP/di < 0. Thus,there is aninverserelationshipbetween the issue price andthe discount rate used: an increase in the dis-count rate that the entrepreneuruses to evaluatethe presentvalue of the tax deferrallowers thewealth-maximizingriceof the IPO.

Summary f TheoreticalFoundationA model of IPO underpricing,where the degreeof underpricing s dependent upon the issuer'sproportionof retainedownership,applicable axrates on income and capital gains, and capitalgainsdeferralcapability,s presentedabove.Sincethe Tax Reform Act of 1986, capitalgains havebeen taxed at the personaltax rate for tax yearsbeginningafter 1986. Analysis of the necessaryconditions for tax-related underpricingrevealsthat if issuersbase theirpricingdecisions,at leastin part, on the tax treatmentof surplusand/orsubsequentcapitalgainsfroman IPO, then IPOsshouldbe less underpriced ubsequent o the taxreform, and as a result, the average degree ofunderpricing houldbe less. That is, in the pres-ence of special lower capitalgains tax rates, thedegreeof underpricinghouldbe greater.

EMPIRICALANALYSISEmpiricalanalysesof IPO underpricing enerallystudy the initial excess returns; i.e. returns inexcess of the marketaveragerate of return.In

particular,he IPO return s measured romofferdate to date of firsttrade in the secondarymar-ket, and the IPO return s adjustedby the returnon the marketindex.This initial excess (market-adjusted) eturn or an IPO is defined n the samemanner as McDonald and Fisher (1972), Blockand Stanley (1980), Neuberger and LaChapelle(1983),and others.Althoughthe model does notexplicitlyconsiderthe risk of an individual tockissue, marketeffects on the returnsof IPOs aretaken into considerationas noted by McDonaldand Fisher(1972)and Beattyand Ritter(1986).

TheDataTwo basic, desirable features of the stocks in-cludedin the samplewere identifiedpriorto datacollection. These are (1) that the issues be ofinitial,stock-onlyofferings,and(2) that the firms'offeringsbe of variedsize. The latterrequirementis necessarysince there may be differingunder-pricing effects due to firm size. As Stoll andCurley (1970), Logue (1973), Bear and Curley(1975),Blockand Stanley 1980)andRitter (1984)hypothesized, ize may influencethe risk of theissue in that largerfirmsmaybe betterknowntothe financialmarketspriorto the IPO, and maybe less riskythan smallerfirms. OTC stocks areused in this studyso that characteristicshat arerestrictedby other stock exchanges,such as firmsize, are not artificiallyimited.The first desirable haracteristic'istedaboveisalso important.Common stocks issued in unitofferingsor mixeddebt-equityofferingswere notincludedin the sample since the returnprocessfor them may reflect the behaviorof the otherfinancial nstruments n the offering,and not ofthe commonstock itself. Consequently, he dataconsist of OTC common stocks that were notoffered with warrantsor in mixed debt-equityofferings.The data examinedfor this paper consist of1308IPOs of firms addedto the list' of Standard& Poor's OTCDailyStockQuotations. irst-tradeprices were gathered from the same source aswere levels of the 'OTC IndustrialIndex'. IPOofferingpricesweregathered rom Moody'sOTCIndustrialManuals.Firm characteristics uch asthe date of incorporation,pre-offertotal assets,long-termdebt, IPO issue size, underwriter, ndthe date of offer were gatheredfrom the same

7/30/2019 2487752

6/10

UNDERPRICING OF IPOS 557Moody'ssource.AAA and BBB monthly ndexesof yields were gatheredfrom the monthlyFederalReserveBulletins, long with monthlyobservationson the CPI. The data span a ten-yearperiod fromJanuary1980 throughDecember 1989.

Testing the Distribution of Excess ReturnsThere is a strikingabsence in the IPO underpric-ing literature of tests of the specific distributionfor excess returns.As noted by Block and Stanley(1980), however,market-adjustedeturnsmay notbe normally distributed.Student's t-test, there-fore, may not be an appropriate est statistic forinitial excess returns.

Given that the mean initial excess return forthe entire sample is 15.97%,with a standarddeviationof 44.51%,andthatthe median s 3.70%,the sample distributionof initial excess returnsexhibitsright-skewness;herefore the populationmay not be normallydistributed.A chi-squaregoodness-of-fit test (see Freund, 1971, p.338;Johnston, 1972, p. 426; Anderson et al., 1981,p.322) of the null hypothesis hat the excess returnsare normally distributed was performed. Thecomputedx2 measure or the samplewas 1702.5.Since, at a 1%level of significance, or a samplesize of 1308, the critical value for the x2 is1443.5, he null hypothesisof normality annot beacceptedat the 1%level of significance.The rejection of the normality hypothesis isconsistent with the findingof sample right-skew-ness. Nonparametricests on shifts in the excessreturns are appropriate ince the null hypothesisof normallydistributed excess returns was notaccepted at the 99% confidence level. Riskingredundancy,however,both parametricand non-parametric ests are presentedbelow.

Testing the Tax Change Effect: ParametricTestsPriorto January1987,capitalgainson assets heldfor at least 6 months were taxed at rates lowerthan ordinary ncome tax rates.The Tax ReformAct of 1986eliminated his practiceand,for eachindividual nvestor,made capital gains taxable atthe investor'sordinaryncome tax rate.For exam-ple, an asset purchasedon the last day of June1986 and sold for a gain on the last day ofDecember 1986 would have been held for 6months and would have qualified or preferential

capital gainstax treatment.If the asset had beenpurchased after the last day of June 1986, theholding periodwould not havebeen long enoughto qualifyfor preferentialcapitalgains tax treat-ment. If the assethadbeen sold after the last dayof December1986, the TaxReform Act requiredthe capital gain to be taxed at the investor'sordinaryncome tax rate.In order to test the effect of the Tax ReformAct on excess returns of IPOs, the data set of1308 returnswas divided into the 1021 observa-tionspriorto and including he first six monthsof1986 and the 287 observations ubsequent o June1986. The mean excess returnfor the first subpe-riod is 18.17%; he standarddeviation s 49.36%.The mean excessreturnfor the second subperiodis 8.11%; he standarddeviation s 17.08%.The t-statisticfor subperiodone is 11.76, with1020 degrees of freedom. The t-statistic in thesecond subperiod is 8.03, with 286 degrees offreedom.Hence, the null hypothesisof no under-pricingin either subperiodseparatelycannot beaccepted at anyreasonableconfidence evel.The tax-basedmodel of underpricing uggeststhat the mean excess return in subperiodoneshould be greaterthan the mean excess return nsubperiodtwo. To test the null hypothesisthatthe mean excess return n subperiodone was lessthan or equal to that of subperiod wo,a pooled-variance -statistic or the difference n the meanswas computed.This t-statisticwas 5.46. Since thecriticalvalue for the t is 2.326 at a 99% confi-dence level, the null hypothesisthat the meanexcess return in subperiodone is less than orequal to that in subperiodtwo cannot be ac-cepted.The parametrictests, then, provide evidencethat althoughexcessreturns existedover the en-tire decade of the 1980s, the degree of excessreturnsdecreasedafter the enactmentof the TaxReform Act of 1986.This is consistentwith thepredictionsof the tax-basedunderpricingmodel.

Testingthe TaxChangeEffect:NonparametricTestsIn order to test via nonparametricmethodswhetherunderpricing xistedin the two subperi-ods a pair-wise ign test was computed.The per-centage change in the IPO price (offer price toprice of firsttrade)was pairedwith the percent-age change in the OTC IndustrialIndex that

7/30/2019 2487752

7/10

558 M. A. RESIDE, R. M. ROBINSON, A. J. PRAKASH AND K. DANDAPANIcorresponded to the same time period. If thepercentagechange in the IPO price was greaterthanthatof the Index,then a '+' wasrecorded;fit was less than the Index then a' - ' was recorded.See Conover 1980, p. 124), Freund 1971, p. 344),or Anderson et al. (1981, p. 418) for explanationsof this test.The numberof positiveobservations n the firstsubperiodwas 632, and n was 871. The criticalvalue for to.01was 470 (zo.01= 2.326 for a = 1%).Since632 > (871 - 470),the null hypothesis hatno underpricingxists in this subperiod annot beacceptedas the 99%confidence evel.The numberof positiveobservations n the sec-ond subperiodwas 185 and n was 143. The criti-cal value for to01was 240 (zo01= 2.326 for a =1%).Since 185> (243 - 140), the null hypothesisthat no underpricing xists in this subperiodalsocannot be acceptedat the 99%confidence level.The Mann-Whitney U-test is a nonparametricmethod used to test the hypothesis hat two sam-ples are drawn from identical populations.SeeFreund(1971,p. 347) and Conover 1980, p. 216)for statisticaldetail.1In this test the two samplesare arrangedointly as thoughthey compriseonesample,andranksareassigned, argest o smallest.Tied observationsare assigned the mean of thetwo spannedranks.The sum of the ranksof oneof the samples is then comparedto a criticalvalue in order to test the hypothesis.The sum of the ranksof the first subperiod s676,535.The Mann-Whitneytest statistic,whichis defined by a standardnormal z-statistic, is26.10. The null hypothesisthat the samples aredrawnfrom identicaldistributions annot be ac-cepted at any meaningful level of significance.The alternativehypothesisthat the underpricingis greater for the first subperiodthan for thesecond cannot be rejected.

MultivariateAnalysisNumerous multiple regressionanalyses of IPOexcess returnshave been presented n the litera-ture. The independentvariables ncludedin theregression,however,have varied greatly.Logue(1973)includedthe numberof IPOs offered dur-ingthe issuingmonth,a Departmentof Commerce'Diffusion Index of Common Stock Prices', adummyvariable to indicate whether or not the'speculative' abel was requiredby the SEC, andcash and noncash compensation ncluded sepa-

rately.UnadjustedIPO returnswere used as thedependent variable. Separate regressions werecomputed for ranked versus nonrankedunder-writers.2For the entiresample,only the diffusionindex andthe total value of the issue were foundto have regressioncoefficients that are signifi-cantly differentfromzero.Bear and Curley 1975)includedthe age of thefirm,the value of the issue, the precedingyear'searnings, the percentage of cash compensationand the firm'sexpost beta, as explanatory ari-ables. UnadjustedIPO returnswere the depen-dent variable.The regressioncoefficientsfor thefirm's ex post beta and noncash compensationwere significantly ifferent romzero.The expostbeta inclusion s of questionable heoreticalvalid-ity, particularly or the purpose of this study,since it could not havebeen knownapriori bythefinancialmarketsgiven that the issues are IPOs.Tinic (1988) included the reciprocalof the of-feringvalue,the natural og of the offeringvalue,and dummy variables for ranked versus non-rankedunderwriters nd for issuing n the monthof January.The dependentvariablewas market-adjustedreturns.Only the coefficientfor the re-ciprocal of the offering price was significantlydifferentfromzero.Beatty andRitter(1986) included he reciprocalof the offering price and an underwriter restigeindex as independentvariables,along with thestated number of uses for the funds raised as aproxyfor uncertaintyconcerningthe returns tothe investor. The latter variable was found tohave no effect. The coefficient for the reciprocalof the offer pricewas significantly ifferent fromzero. The dependentvariablewas the market-ad-justedrate of return.The regressionsbelow use the initialmarket-ad-justed rate of returnas the dependentvariable.The independentvariables nclude the firm'sage,the value of the firm'sassets,its debt/asset ratio,the issue value, the underwriter's ankingas in-dexed in Carter and Manaster (1990) and adummyvariableof '1' for issues after June 1986and'0' for those issued before.3The realvalue ofthe firm'sassets and issue size are also included.In addition, because the pricing model pre-sented above shows that IPO prices should bedependent upon interest rates via discount fac-tors, an attemptwas made to use BBB andAAAbond indexes as independentvariables. Becauseof the general decline in interest rates over the

7/30/2019 2487752

8/10

UNDERPRICING OF IPOS 559later years of the 1980s,however,these rate in-dexes were highly correlated with the dummyvariable.The decrease in marginal ax rates thatwere implemented due to the Tax Reform Actwould cause the supplyof loanable funds to in-crease and rates to fall. This would result fromthe increase in after-taxrates. The correlationcoefficient between BBB yields and the dummyvariablewas - 0.72,andbetweenAAA yieldsandthe dummyvariables t was - 0.71. Using eitherof these rates along with the dummyvariablewouldcause severemulticollinearity roblemsandmake it impossibleto discern the effect of thedummyvariable from the rate indexes. Hence,the indexes were omitted from the regressions.4

For the purpose of examining possible multi-collinearityproblemswith the independentvari-ables,Table 1 presentsthe correlationcoefficientmatrixof the variablesused in the regression.Asshown,the dummyvariablehas a correlationco-efficient of 0.20 with the underwriterrankingvariable.This slight degreeof collinearityhas thepotential to partially obscure the influence ofeach upon the dependentvariable. The dummyvariable also has a slight correlation with thevalue of the assets and the issue value. Thesecorrelationsare probably,in part, due to theinflationaryrend of the 1980swhere the value ofthe assets and the issue value increased n nomi-

nal terms over the decade. Deflating these vari-ables by the CPI lowered their correlations.Themore serious correlationproblem exists betweenthe underwriter ankingvariableand the value ofthe assets and issue. Deflating these variablesloweredthe correlations lightly.The five regressionscomputed and reported nTable 2 show that the underwriter anking,debt-to-assetratio, dummyvariableand the age of thefirm had coefficients that were significantlydif-ferent fromzero at highprobability evels.By themagnitudeof the coefficients, he debt/asset ra-tio had the largest impact on the IPO returns.The returns are measuredas percentages,hencea change in the debt/asset ratio from 0.5 to 0.6has an impactof - 2%on the rateof return.Thisfinding is consistent with James (1992). Thedummyvariable has the second-largest mpact,with regressioncoefficients of about -6, whichmeans that the mean excess return for IPOs fellby 6% after the Tax Reform Act was imple-mented. The underwriterrankingvariable andthe third-largestmpact,with coefficientsof about-3. This indicates that a ranking of 9 causesexcess returns o be, on average, owerby 27%ascompared o underwriter ankingsof 0. In threeof the regressions,however,the dummyvariablehad significance evels of 98% rather than the

Table 1. CorrelationCoefficientsof RegressionVariablesRANK TA D/A DUM AGE VAL RET

RANK 1.000TA 0.328a 1.000D/A 0.097a 0.181a 1.000DUM 0.201a 0.166a 0.034 1.000AGE 0.131a 0.091a 0.016 -0.007 1.000VAL 0.466a 0.452a 0.061b 0.167a 0.053b 1.000RET -0.230a -0.081a -0.107a -0.071a _0.090a -0.067a 1.000RVAL 0.465a 0.065a 0.114a 0.082a 0.098aRTA 0.2622 0.105a 0.126a 0.088a 0.072aaSignificantlyifferent rom zero at 99% confidence evel.bSignfficantldifferent romzero at 95% confidence evel.RANK:Prestigeranking f the managingunderwriter, to 9, accordingo Carterand Manaster 1990).TA: Total pre-offerassets of the IPO firm $ millions).D/A: Pre-offerdebt-to-asset atio of the IPO firm.DUM: Dummyvariable o '0' prior o 7/86, or '1' after 6/86.AGE: Age of the IPOfirmmeasuredn whole years,0 for firmsof age less than oneyear.VAl: Total valueof the issue ($ millions).RET: IPOinitial excess return %), offer price to firsttradeprice.RVAL:Value of the issue dividedby the CPI ($ millions).RTA: Value of assets dividedbythe CPI($ millions).

7/30/2019 2487752

9/10

560 M. A. RESIDE, R. M. ROBINSON, A. J. PRAKASH AND K. DANDAPANI

Table 2. OLS RegressionCoefficientsaRANK -3.04 -3.03 -3.13 -3.18(4.44)b ( 4.43)b (-5.01)b (-4.66)bTA -0.04 -0.06 -0.08(-0.17) (-0.28) (-0.34)D/A -19.95 -19.94 -19.96 -20.42

(-3.68)b (-3.68)b (-3.68)b - 3.77)bDUM -6.12 -6.02 -6.10 -10.07

(- 2.06YC (- 2.01)Y (-2.05)Y( 3.40)bAGE -0.48 -0.48 -0.48 -0.47

(-3.62)b (-3.62)b (-3.63)b - 3.59)bVAL -0.32 -0.45(-0.37) (-0.53)RVAL -13.82(-0.18)RTA -90.32(-0.34)R2 0.05 0.05 0.05 0.04 0.01

t-statistics re in parentheses.bThe dependentvariable s the market-adjustedeturn RET).The regression oefficient s significantly ifferent romzero at the 99%confidence evel.CThe egression oefficient s significantly ifferent romzero at the 98%confidence evel.

99% probability evels for the other significantvariables.The negative signs of the coefficients for thedummyvariable, he rankingvariableandthe ageof the firm are all consistent with theoreticalpredictions.Since the Tax Reform Act loweredthe capitalgainstaxrates,it is expectedthat IPOreturns should decrease,as predictedby the tax-based pricing model. This is consistentwith thenegativecoefficient.The use of higher-prestige nderwriters houldlower the rate of returnof the issue in the after-market.That is, as suggested n previousstudies,IPOs should be underpriced o a lesser extent ifhigher-ranked nderwriters re used by the issu-ingfirm.The negativeregressioncofficient s con-sistent with this prediction.An increase in thefirm'sage should also lower the IPO return sincemore shouldbe knownby the financial marketsabout the olderfirms;hence; uncertaintys lowerfor older firms.The negativecoefficienton the debt/asset ratiois not, however, clearly explained by theory. Itmightbe expectedthat the greaterthe debtratio,

the greater the uncertaintyas to the future re-turns to the IPO issue holder,and, therefore,thegreater the returnrequiredby the financialmar-kets. It could be, however,that those IPO firmsthat previouslyhad issued debt are better knownto the financial markets, possibly favorablyaf-fecting the acceptanceof their publicequity ssue.This could explainthe negativecoefficient.

SUMMARYAND CONCLUSIONSA theory of tax-based IPO underpricings pre-sented above. An entrepreneurialwealth-maxi-mizing model is given that shows that deferredcapital gains taxes, combinedwith currentlydueordinaryncome taxes, couldpartiallyaccountforIPO underpricing.The model explicitlypredictsthat an increase in the capital gains tax rate,ceterisparibus,shouldresult in a lowerdegree ofunderpricing.The Tax ReformAct of 1986providesan oppor-tunityto test this tax-basedunderpricing ypothe-sis since the Tax Reform Act raised the capital

7/30/2019 2487752

10/10

UNDERPRICING OF IPOS 561gainstax rate. It is shown that the distribution fexcessreturnsdid shift aboutthe implementationdate of the Act, and in the directionpredictedbythe model.This shift is verifiedby both paramet-ric and nonparametrictatisticalmethods.In ad-dition, regressionanalysesthat controlfor otherpossibleunderpricingactorsandthatallowed heIPO returns to shift about the Reform Act'simplementationdate via a dummyvariablecon-firms the predictedtax effect. The empiricalevi-dence examinedis therefore consistentwith thepredictionsof the tax-basedpricingmodel.

AcknowledgementsWe would like to thank an anonymousreviewerfor theconstructive ommentswe receivedon the earlierversionsofthe paper.Any remaining rrorsare our sole responsibility.

NOTES1. The Mann-WhitneyU-test is appropriateor sam-ples of unequalsize.2. Amongothers,Logue (1973)and Tinic (1988)haveprovideddescriptionsof 'rankedversus nonranked'(or 'prestigiousversus nonprestigious') nderwritergroupings.Briefly, the classificationsoosely followthe description ivenby Hayes(1971),whereunder-writerslisted at the top of a tombstonehave hadhigherrank han thoseat the bottom.Higherrank ssynonymouswith higher stature or prestigein theunderwriting ommunity.3. The inclusionof the debt/asset ratio shouldprovidemore information o the regression han a dummyvariable for the presence of debt, as employedbyJames 1992).James ounda significant nd negativerelationshipbetween IPO underpricingand thepresenceof debt in the firm's apitalstructureat thetime of issue. A value of '1, was assignedto thedummy ariable or firmswith debtand '0' for firms

withoutdebt. The finding uggests hat if a firmhaspreviouslyssued debt, the IPO is underpricedo alesser extent than if the firmhas not issued debt.4. By orthogonalizingcollinear variables, some re-searchershave attemptedto discerntheir relativeinfluences(see Singhet al., 1991).This assignsonevariable to the orthogonalizedresidual, however,and the choiceof assignments arbitrary.t doesnotseparate he true explanatory owerof each.

REFERENCESD. R. Anderson, D. J. Sweeney and T. A. Williams(1981) Introductiono Statistics:An Applications p-

proach,St Paul: West Publishing.R. M. Bear and A. J. Curley 1975)UnseasonedequityfinancingJoumalof Financial ndQuantitativeAnaly-sis, 311-25.R. P. Beattyand J. R. Ritter(1986)Investmentbanking,reputation,and the underpricingof initial publicofferingsJoumalof FinancialEconomics,XV,213-32.S. Block andM.Stanley 1980)The financial haracter-istics and price movement patterns of companiesapproachinghe unseasoned ecuritiesmarket n thelate 1970's FinancialManagement,0-36.R. Carterand S. Manaster 1990)Initialpublic offer-ings and underwriter eputation Joumalof Finance,XLV,September,1045-64.W. J. Conover 1971)PracticalNonparametrictatistics,New York:John Wiley.K. Dandapani,R. Dossani, A. J. Prakashand M. A.Reside (1992)Personal axes andthe underpricingfinitial public offerings Managerialand DecisionEconomics,July-August,279-286.J. E. Freund (1971) Mathematical tatistics, nd edn.,EnglewoodCliffs, NJ:Prentice-Hall.S. L. HayesIII (1971)Investmentbanking:powerstruc-ture in flux HarvardBusinessRemiew,March-April,136-52.C. James (1992) Relationship-specificssets and thepricingof underwriter ervices The Joumal of Fi-nance,XLVII,5, 1865-85.J. Johnston 1972)EconometricMethods, rd edn.,NewYork:McGraw-Hill.D. E. Logue (1973) On the pricing of unseasonedequity issues: 1965-1969 Joumal of FinancialandQuantitativenalysis,91-103.J. G. McDonaldand A. K. Fisher (1972) New-issuestockpricebehavior TheJoumalof Finance,97-102.B. Neubergerand C.A. LaChapelle 1983)Unseasonednew issue price performance on three tiers:1975-1980FinancialManagement,3-8.J. R. Ritter (1984) The 'hot issue' market of 1980Joumalof Business,LVII,2, 215-40.

A. K. Singh,R. Cowanand N. Nayar 1991)Underwrit-ten calls of convertiblebonds Joumalof FinancialEconomics,XXIX, 173-96.H. R. Stolland A. Curley 1970)Smallbusinessandthenew issues market for equities Joumal of FinancialandQuantitativenalysis,V, 309-22.S. M. Tinic(1988)Anatomyof initialpublicofferingsofcommon stock The Joumal of Finance, XLIII, 4,789-822.