pdf viewer control without acrobat reader installed c# : Add picture to pdf document software SDK cloud windows wpf html class pdf_CC250-part2625

MANAGEMENT SCIENCE
Vol.51,No.7,July2005,pp.1046–1059
issn0025-1909eissn1526-55010551071046
inf
orms
®
doi10.1287/mnsc.1040.0358
©2005INFORMS
The Financial Impact of ISO 9000 Certification in
the United States: An Empirical Analysis
CharlesJ. Corbett
AndersonSchoolofManagement,UniversityofCalifornia,LosAngeles,California90095-1481,
charles.corbett@anderson.ucla.edu
MaríaJ.Montes-Sancho
UniversidadCarlosIIIdeMadrid,Madrid126–28903,Getafe,Madrid,Spain,mmontes@emp.uc3m.es
DavidA.Kirsch
R.H.SmithSchoolofBusiness,UniversityofMaryland,CollegePark,Maryland20742,dkirsch@rhsmith.umd.edu
T
he ISO 9000 series of quality management t systems standards, introduced in 1986,has been adopted at
over 560,000 locations worldwide.Anecdotal evidence e suggests that firms can achieve internal benefits
suchas quality y or productivity improvementsor that certification canhelp firmsmaintainor r increase their
market share,orboth.Othersarguethat thestandardistoo genericto cause performance improvement but
canbeseenasasignalofgoodmanagement.Inthispaper,wetrackfinancialperformancefrom1987to1997
ofall publiclytradedISO9000certifiedmanufacturingfirmsinthe UnitedStateswithSIC codes2000–3999,
andtest whether ISO O 9000certificationleadsto productivity improvements,market t benefits, andimproved
financialperformance.Weemployevent-studymethods,matchingeachcertifiedfirmtoacontrolgroupofone
ormorenoncertifiedfirmsinthesameindustrywithsimilarprecertificationsizeand/orreturnonassets.We
findthat firms’decisiontoseektheirfirst ISO9000certificationwasindeedfollowedbysignificantabnormal
improvementsinfinancial performance,thoughtheexact timing andmagnitudeofthiseffect dependonthe
specificationofthecontrolgroup.Threeyearsaftercertification,thecertifiedfirmsdodisplaystronglysignificant
abnormalperformanceunderallcontrol-groupspecifications.Thedegreetowhichthepreciseresultsvaryacross
control-groupspecificationsindicatesthat event studiesshould alwaysinclude extensive sensitivity analysis,
forinstancematchingbysizeandperformanceseparatelyandjointly,usingbothsinglefirmsandportfoliosas
controls.
Keywords: ISO9000;qualitymanagement;standards;financial;empirical;eventstudy;Compustat
History: AcceptedbyWilliamS.Lovejoy,operationsandsupplychainmanagement;receivedJune21,2002.
Thispaperwaswiththeauthors11monthsfor3revisions.
1. Introduction
TheISO9000seriesofqualitymanagement systems
standardsisanexceptionallywidelydiffusedmanage-
mentpractice.ByDecember2001,ithadbeenadopted
byover560,000facilitiesin159countries;over38,000
of those facilities are located in the United States
(ISO 2003). Despite this, considerable uncertainty
existsamongpractitionersandscholarsaboutthetrue
effects of the standard, given its generic and mini-
mallyprescriptivenature.ReimannandHertz(1994)
explain ISO 9000’s focus on conformance and how
it differs from the Baldrige award’s much broader
focus on competitiveness. Moreover, implementing
ISO 9000 is not without cost. Besides the fees for
auditors and consultants, the process generallycon-
sumes considerable e employee time andeffort. Espe-
ciallyasISO9000was thefirst standard ofitstype,
many firms had to develop, implement, and docu-
menttheirqualitymanagementsystemfromscratch.
Remarkably, though, there isno hard evidence to
dateonthefinancialeffectsofISO9000certification.
LandmarkstudiesbyHendricks and Singhal (1997)
and Easton and Jarrell (1998) show that total qual-
itymanagement (TQM) leadsto improved financial
performance,butthe philosophyunderlyingTQMis
quite different from (though not inconsistent with)
that ofISO 9000. TQMis alooselydefined, holistic
set of principles and tools aimed at, amongothers,
encouragingcontinuousimprovementandprevention
of defects; the premise of ISO 9000 is that well
definedanddocumentedproceduresimproveconsis-
tencyofoutput. Degree of TQM implementation is
difficult to assessobjectively:HendricksandSinghal
(1997)usequalityawardstosignalthepresenceofan
effective TQMprogram, and Hendricksand Singhal
(2001b)findthattheexactnatureofthequalityaward
influences the extent to which recipient firms ben-
efit from it. Easton and Jarrell (1998) use in-depth
1046
Add picture to pdf document - insert images into PDF in C#.net, ASP.NET, MVC, Ajax, WinForms, WPF
Sample C# code to add image, picture, logo or digital photo into PDF document page using PDF page editor control
how to add a jpeg to a pdf file; add image to pdf online
Add picture to pdf document - VB.NET PDF insert image library: insert images into PDF in vb.net, ASP.NET, MVC, Ajax, WinForms, WPF
Guide VB.NET Programmers How to Add Images in PDF Document
adding a png to a pdf; how to add photo to pdf in preview
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
ManagementScience51(7),pp.1046–1059,©2005INFORMS
1047
questionnaires to assess degree of TQM implemen-
tation. ISO 9000 certification, on the other hand, is
unambiguous,duetothethird-partyauditsinvolved.
Anecdotal evidence suggests that some firms truly
take the underlyingprinciplesofthe ISO9000stan-
dards to heart, while others only do the mini-
mumnecessarytopassthe audits. Indeed,Andrews
etal.(2001)find noticeabledifferencesbeweenaudi-
tors’ interpretationsofthe e ISO14001standards, the
environmental management systems counterpart to
ISO9000.Someoftheareasofvariationwerespecific
toISO14001,whileothersarelikelytoapplyequally
toISO9000,suchasvariationsinauditortrainingand
experience.Inthisstudy,wefocusonlyontheeffects
ofcertification,ignoringthedepthofimplementation.
Most scholarly work so far on the benefits of
ISO9000hasreliedonsurveys;certified firmsoften
claim to have obtained benefits from adopting ISO
9000,butthesebenefitsareself-reported.Inthispaper,
we combine Compustat data with a database e ofall
ISO 9000 certifications in the United States in the
manufacturingsector,withSICcodes2000–3999.The
resultis adatabase e of7,238 firmsfor the 1987–1997
period. We perform a range of sensitivity analyses
to explore the impact of different matching meth-
ods.Wematchcertifiedfirmstoindividualnoncerti-
fiedfirmsandtocontrolgroupsofnoncertifiedfirms.
We use several criteria to select control firms: pre-
certification size, returnon assets,andboth. We use
variousparametric and nonparametrictest statistics.
Inallcases,wefind thatthree years after their first
certification, firms experienced strongly significant
abnormalperformance,thoughthe exact timingand
magnitudeofthe effectsvarywithdifferent control-
groupspecifications.
In §2, wereviewrelevant literature related to ISO
9000andTQM.Section3developsthetheoreticalrela-
tionshipsbetweenISO9000certificationandfinancial
performance. Section 4describesour data. Section5
describestheevent-studymethodsusedandpresents
ourmainresults.We discussthese andoursensitiv-
ityanalysisin§6.Section7offersourconclusionsand
suggestionsforfutureresearch.
2. Literature
InthissectionwereviewliteratureonISO9000certifi-
cationandonTQM.MostliteratureonISO9000either
focuses on adoption or uses self-reported survey-
basedperformance measures.SurveysbyTerziovski
et al. (1997) and Singels et al. (2001) find no link
between ISO 9000 and organizational performance
among Australian and Dutch firms, respectively,
though Terziovski et al. (2003) do find that higher
reported business performance is positively associ-
ated with a broader set of motivations for seeking
ISO9000certification.Casadesúsetal.(2001)findthat
Basquecompaniesreportsubstantialoperationaland
financialbenefitsfromcertification.
UsingCompustatdata, Andersonetal.(1999)find
that U.S. firms with higher exports to Europe are
morelikelytoseekISO9000certification,andCorbett
(2004),usingdatafromamultinationalsurvey, finds
that exports drive early certifications; Guler et al.
(2000),Mendel(2001),andCorbettandKirsch(2001)
examinedriversofglobal diffusionofISO9000and
ISO14000.Adams(1999)findsthatISO9000certifica-
tioninNewZealandincreaseswithfirmsize,Tobin’s
Q, and market segmentation. Naveh and Marcus
(2004), using a detailed survey of ISO 9000 in the
United States,find that “goingbeyond”the require-
mentsofthe standardincreasesitsvaluetothe firm.
Terlaak and King (2005a) argue that ISO 9000 is a
signallingmechanism and findssalesgrowththat is
partly consistent with that view; Terlaak and King
(2005b) also find evidence that institutional theory
contributesto explainingadoption. King and Lenox
(2001)andRusso(2001)findsome evidence thatISO
9000andISO14000certificationrespectivelyimprove
firms’ environmental performance, suggesting that
bothstandardsdoalsohaveanimpactonproduction
practices.
Severalauthorshaveexaminedthefinancialeffects
of implementing TQM programs. Hendricks and
Singhal(1996,1997,2001a)useeventstudiestoshow
that winning a quality award increases short-term
andlong-termstockpriceandoperatingperformance.
Hendricks and Singhal (2001b) find that the extent
of financial benefits arising from TQM depends on
firmcharacteristicsandonthetypeofqualityaward
involved. The event studies in Easton and Jarrell
(1998) also show a significant improvement in per-
formance after implementing TQM. Powell (1995)
concludes that the tacit resources often associated
withsuccessfulTQMimplementationimproveperfor-
mance,ratherthantheTQMtoolsthemselves. Ittner
etal.(2001)reportthatproactiveinvestmentsinpre-
vention do lead to lower nonconformance quality
costs.
Bycontrast, veryfewauthorshave explicitlymea-
sured the financial effects of ISO 9000 certification.
Event studies by Lima et al. (2000) and Martínez-
Costa and Martínez-Lorente (2002) find no effects
in Brazil and Spain, respectively, while the Spanish
event studybyNicolauandSellers(2002)onlyfinds
a positive e stock price effect on the e day of certifica-
tionitself.DockingandDowen(1999)findthatsmall
firmsintheUnitedStatesexperiencedpositive stock
market reaction to the announcement of their first
ISO 9000 certification, but that larger firms’ stock
pricedidnotrespond.SimmonsandWhite(1999)find
thatISO9000certificationisassociatedwithsuperior
financialperformanceintheUnitedStateswithinSIC
C# TIFF: How to Insert & Burn Picture/Image into TIFF Document
Support adding image or picture to an existing new REImage(@"c:\ logo.png"); // add the image powerful & profession imaging controls, PDF document, tiff files
add jpg to pdf; add image field to pdf form
VB.NET TIFF: How to Draw Picture & Write Text on TIFF Document in
Dim drawing As RaterEdgeDrawing = New RaterEdgeDrawing() drawing.Picture = "RasterEdge" drawing powerful & profession imaging controls, PDF document, tiff files
how to add image to pdf reader; adding image to pdf form
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
1048
ManagementScience51(7),pp.1046–1059,©2005INFORMS
code36(electronicandotherelectricalequipmentand
components, except computer equipment), but have
insufficient longitudinal data to indicate the direc-
tion ofcausality. Heras et al. (2002)find that more
profitable firmsinthe Basque regionaremore likely
to seek ISO 9000 certification, but that theydo not
becomemoreprofitableasaresult.Byusingthemuch
largerU.S.database,focusingonlong-termoperating
performance,andmatchingfirmsbasedonprecertifi-
cationperformance, weareable to detectsignificant
andpersistenteffectsofcertification.
3. TheoryDevelopmentand
Hypotheses
Our theoretical basisisanalogoustothat commonly
foundintheTQMliterature,forinstanceinHendricks
andSinghal(1997)andEastonandJarrell(1998).The
startingpointisthehypothesisthat ISO9000certifi-
cationhasapositiveeffectonprofitability,regardless
ofthe underlying mechanism, where profitabilityis
commonlymeasuredbyreturnonassets(ROA:oper-
atingincome/assets)orreturnonsales(ROS:operat-
ingincome/sales),bothbeforedepreciation.
Hypothesis1. ISO 9000 certification leads to an
increaseinreturnonassets(ROA)andinreturnonsales
(ROS).
A corollaryof this hypothesis is that certification
contributestothestockmarket’svaluationofafirm’s
intangibleassets,measuredbyTobin’sQ;afirmwith
the same physical assets but “better management”
should have ahighermarketvaluation. Tobin’sQis
defined as“theratio ofthe market valueoftheout-
standingfinancial claimson the firm to the e current
replacementcost ofthefirm’sassets”(Lewellenand
Badrinath1997,p.78),andmeasuresthealternative-
usevalueofthefirm’sassets.Forthehorizonwecon-
siderhere,Tobin’sQcanalsobethoughtofasaproxy
forstockpricemovement.
Hypothesis2. ISO 9000 certification leads to an
increaseinTobin’sQ.
Exactly how is ISO 9000 certification linked to
a firm’s financial performance? The academic and
practitionerliteraturesproposeseveralpossiblelinks.
ISO9000mayimprove afirm’sinternal procedures,
and hence enhance performance by reducing cost.
Alternatively, ISO9000certificationmayhelpafirm
retain or increase market share, and hence improve
performancebyincreasingsalesorrevenues.
3.1. ProductivityModel:ISO9000Leadsto
InternalImprovements
Implementing ISO 9000 forces significant discipline
on the firm: it has to design procedures to ensure
qualityisconstantlymeasured,procedurestoensure
appropriate corrective action is taken whenever
defects occur, etc. As a result, defect rates should
decrease and defectsshouldbecaughtearlierallow-
ingthemtobecorrectedatlowercost.Thisdiscipline
can also identify current practices that are obsolete
or counterproductive. The procedures must be suf-
ficientlywell defined that theycanbe documented.
Once documented, theybecomeeasier to transfer to
newemployees,sothefirmbecomeslessdependent
on tacit knowledge residing in individual employ-
ees (Mukherjee et al. 1998). Elmuti and Kathawala
(1997)reportthatISO9000increasesproductivityby
improving employee morale. Through all of these
mechanisms, manufacturing costs should decrease
whena firm starts implementing ISO 9000. We e use
cost ofgoodssold dividedbysales (COGS/SALES)
to measure such improvements in manufacturing
productivity.
Hypothesis3. ISO 9000 certification leads to lower
COGS/SALESratios.
A broader view of a firm’s overall productiv-
ity includes customer service processes, the cost
of the certification, etc., which fall under selling,
general, and administrative (SG&A) expenses. We
do not report the broader cost measure (COGS+
SG&A)/SALESseparatelybecauseitislinkedtoROS
through1−(COGS+SG&A)/SALES.
3.2. MarketModel:ISO9000Leadsto
MarketBenefits
How does ISO 9000 certification help in the mar-
ket?Certificationmayincreasethelikelihoodofwin-
ning a contract, or it may even be a requirement.
Alternatively, certification mayhelp firmsto reduce
pricepressure.Inbothcases,thevalueofsaleswould
increase. Thiscould occur after certification, though
the improved proceduresimplemented prior to cer-
tificationcould also alreadypositivelyimpact sales.
Hence,weuserelativegrowthinSALES,definedby
SALES
t
−SALES
t−1
/SALES
t−1
, asone ofour exter-
nal performance measures. To determine whether
sales growth reflects firm growth or faster asset
turnover, we also consider SALES/ASSETS, follow-
ing Hendricks and Singhal (1997) and Easton and
Jarrell(1998).
Hypothesis4. ISO 9000certificationleads to higher
relative SALES growth and to increased asset turnover
(SALES/ASSETS).
4. Data
Thisstudyisbasedontwolargedatasets.WorldPre-
ferred kindlyshared their database (then known as
the GlobusRegistry)ofISO9000certificationsinthe
C# Word - Paragraph Processing in C#.NET
Add references: CreateParagraph(); //Create a picture for para IPicture picture = para.CreatePicture(imageSrcPath); //Save the document doc0.Save
add a picture to a pdf; add a jpg to a pdf
VB.NET Image: Create Code 11 Barcode on Picture & Document Using
file, apart from above mentioned .NET core imaging SDK and .NET barcode creator add-on, you also need to buy .NET PDF document editor add-on, namely, RasterEdge
add picture to pdf online; add png to pdf acrobat
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
ManagementScience51(7),pp.1046–1059,©2005INFORMS
1049
UnitedStatesthrough1997,atotalof21,482certifica-
tionsinallSICcodesatthattime.Thisdatabaselisted
thenameandaddressofthecertifiedsite,theparent
company,primaryandsecondarySICcode,anddate
of certification. World Preferred is the e most widely
usedsource ofinformation onISO9000certification
activity in the United States (perhaps jointly with
McGraw-Hill).Althoughwearenotawareofinaccu-
raciesintheWorldPreferredRegistry,weaddressthe
consequencesofpossibleerrorsbelow.
Firmscanhavemultiplecertifications,evenwithin
onesite,whilefinancialperformance ismeasured at
thecorporatelevel;thereexistsnowayofmeasuring
financialperformanceofthatpartofacompanythat
fallswithinthescopeofagivenISO9000certification.
Therefore, followingDockingandDowen(1999)and
SimmonsandWhite(1999),wefocusonafirm’sfirst
certification,astheeffectsofcertification(ifany)will
begintoshowaroundthattime.Forsmallfirmswith
asingle site, the disconnect betweensite-levelcerti-
fications and firm-level performance is not a major
concern, while for large firmswithmanysites, this
disconnect makesour testsmore conservative: ifthe
samplecontainsmanymultisitefirmswithonlyone
certifiedsite,anyeffectsofcertificationmustbestrong
inordertobedetected.Wereturntothispointinour
discussionofthelimitationsofthisstudy.
Ourfirstcertificationoccurredin1990,soweused
CompustatAnnualIndustrialFiledatafor1987–1997.
Mergingthetwodatabaseswasamajorchallenge,as
nouniquecompanyidentifierwaslistedinboth.We
used severalmethodsto merge them: in some e easy
instances, the parent company name and SIC code
in the e certification database matched the e name and
SICcodeofacompanylistedinCompustat.Inmany
cases,companieswithsimilarbutnonidenticalnames
appeared in both databases; we used several auto-
mated proceduresto identifysuchpossible matches
andverifiedthemmanually,usingvariousresources
suchasHoover’sOnline,Google,companywebsites,
SEC disclosures, etc. In all other cases, we verified
whether asite e or parent company listed inthe cer-
tificationdatabase wasinfact a subsidiaryofa dif-
ferently named company in Compustat. Sites with
ISO9000certificationforwhichnone ofthese proce-
duresyieldedamatchinCompustatwerediscarded;
thesearelikelytobeforeignownedorprivatelyheld,
and hence not listed inCompustat. We started with
21,482 certified sites; for 7,598 of these, we identi-
fied matching firms in Compustat, which suggests
that approximately35.4%ofcertifiedsites belongto
publicly traded firms. This percentage corresponds
closelytothatinarecentsurveyofISO9000andISO
14000certifiedsitesinthe UnitedStates(seeCorbett
2004),where 67% of933respondentsreportedbeing
foreignownedorprivatelyheld.
Altogether,mergingthe datatookover two years.
Giventhe scopeandthe manualnature ofthework
involved,twotypesoferrorsaretheoreticallypossi-
ble: firmsthat did receive certification in1990–1997
may be listed as noncertified in our final data set,
andfirmsthatdidnotreceivecertificationmayerro-
neouslybelistedascertified.Inarandomsample of
100firms, we found onlyonefirmlisted asnoncer-
tified for whichwe were eventuallyable to identify
a certification, and no instances offirmsincorrectly
labeledascertified.Thelattertypeoferrorisindeed
highly unlikely, as we only mark Compustat firms
as certified ifwe have apositive e match. Incorrectly
labelingaCompustatfirmasnoncertifiedcouldoccur
moreeasily,butthiswouldonlymakeourtestsmore
conservative, as longasthe e reliabilityofthematch-
ing is not correlated with the interaction between
ISO9000certificationandfinancialperformance.Ifwe
treatsomecertifiedfirmsasnoncertified,butstillfind
asignificanteffect (aswe do),wewould mostlikely
underestimate its significance. Moreover, given the
verylowerrorratewedetected,anysucheffectsare
likelytobeminimal.
AnotherpotentialsourceoferrorsliesintheWorld
Preferred Registry itself, though there are reasons
to believe the e database isaccurate. World Preferred
receives certification information directly from the
registrarsresponsible.RegistrarsandWorldPreferred
wouldclearlywanttoavoidlistingcertificationsthat
have not been awarded. Conversely, one e use of the
World Preferred Registry is to determine the mar-
ketshareofregistrarsoperatingintheUnitedStates;
therefore, registrars will want the World Preferred
Registrytocontainallcertificationsforwhichtheyare
responsible,asotherwisetheirmarketsharewouldbe
underestimated.Similarly,individualfirmswillwant
their certification to be appropriately listed. Even if
some certificationsare not includedinthe database,
thiswouldagainmakeourtestsmoreconservative.
We focused onmanufacturingfirms, i.e., allfirms
withSICcodes2000–3999.Inmostinstances,weselec-
ted control firms usingtwo-digit SICcodesbecause
ofthehigherproportionofdiscrepanciesinthree-or
four-digitSICcodes(GuentherandRosman1994)and
becausethesamplesizeotherwiseshrinksrapidly.We
didseparateSICcode357(computerequipment)from
the other sectorswithinSICcode 35(industrial and
commercial machinery, which includes engines and
turbines,farmandgardenmachineryandequipment,
etc.) as it is far more high-tech innature e and suffi-
cientlylargetotreatseparately.
Of the 7,238 firms listed in Compustat with SIC
codes 2000–3999, 1,103 firms received at least one
ISO 9000 certification during 1990–1997. We elim-
inated 310 observations with insufficient financial
VB.NET PowerPoint: Add Image to PowerPoint Document Slide/Page
clip art or screenshot, the picture will be AddPage", "InsertPage" and "DeletePage" to add, insert or & profession imaging controls, PDF document, tiff files
how to add a jpeg to a pdf; how to add image to pdf in acrobat
VB.NET Image: VB.NET Planet Barcode Generator for Image, Picture &
on Overview. VB.NET Planet Barcode Creator Add-on within Generate Planet Barcode on Picture & Image in VB.NET. In for adding Planet barcode image to PDF, TIFF or
add image to pdf in preview; add png to pdf preview
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
1050
ManagementScience51(7),pp.1046–1059,©2005INFORMS
information. We then matched each firm to a con-
trol firmorportfoliobased onprecertification ROA,
assets,orboth(aswellasindustry);ifnosuitablecon-
trolfirmexistedforanyoneofthematchingcriteria,
theobservationwasdiscarded.Thiswasdonetokeep
the sample constantunderdifferent matchingmeth-
ods, in order to avoid our comparisons being con-
taminated bysample effects.Afterthat, wetrimmed
thedata(followingBarberand Lyon1996), eliminat-
ing afurther one to two outliersat each tailofthe
population. After trimming, we have 554 firms for
whichcontrolfirmsexistunderallmatchingcriteria.
Tables 1a and 1b give descriptive statistics for our
sample.
MeasuringTobin’s Qisnot unambiguous; Chung
and Pruitt (1994) propose approximating Tobin’s Q
by (MVE+PS+DEBT)/TA, where MVE is market
valueofequity,PStheliquidatingvalueofthefirm’s
outstanding preferred stock, DEBT the value e of the
Table1a
SummaryStatisticsfor Certified d and Control Firms s Using
One-to-OneMatchingbyIndustryCombinedwithROAand/or
ASSETS(Yeart−2)
N
Mean
Median St.dev.
Min.
Max.
Certifiedfirms
TOTALASSETSa 554 162932 21882 571709
157
7429300
SALESa
554 164911 25277 554670
343
7955700
ROAb
554
1572 1526
812 −3308
6069
ROSb
554
1328 1252
830 −6044
5293
Tobin’sQc
501
151
104
149
010
1425
COGS/SALESb
554
6303 6545
1572
1757
9480
SALES/ASSETSb 554
12740 12162
4542
1834
39547
ControlfirmsmatchedbyindustryROAandASSETS
TOTALASSETSa 554 166283 19710 626639
132
8634300
SALESa
554 174933 26684 709730
170 11049600
ROAb
554
1566 1516
803 −3214
5816
ROSb
554
1330 1252
1565−29566
4853
Tobin’sQc
449
150
118
122
−083
1158
COGS/SALESb
554
6367 6648
1698
1293
14736
SALES/ASSETSb 554
12708 11865
5420
1087
38225
ControlfirmsmatchedbyindustryROA
TOTALASSETSa 554 168139 7171 974992
023 12130770
SALESa
554 157776 9627 895379
027 11857160
ROAb
554
1571 1531
812 −3214
6282
ROSb
554
1281 1238
1566−29566
4267
Tobin’sQc
438
158
114
141
−007
1158
COGS/SALESb
554
6286 6421
1730
891
14736
SALES/ASSETSb 554
13308 12434
5988
1087
45615
ControlfirmsmatchedbyindustryASSETS
TOTALASSETSa 554 164390 21876 627525
148
8528000
SALESa
554 158892 25420 575514
125
7839400
ROAb
554
1273 1280
1055 −3289
5879
ROSb
554
931 1115
2501−33440
4810
Tobin’sQc
457
137
100
113
−051
936
COGS/SALESb
554
6333 6572
1810
1038
16275
SALES/ASSETSb 554
11961 11423
5362
202
37206
aTOTALASSETSandSALESareinmillionsofdollars.
bROA,ROS,COGS/SALES,andSALES/ASSETSareexpressedinpercent-
ages.
cTobin’sQisadimensionlessmeasure.
Table1b
Summary Statistics for Certified d and Control FirmsUsing
Portfolio MatchingbyIndustryCombined withROAand/or
ASSETS(Yeart−2)
N
Mean
Median St.dev.
Min.
Max.
Certifiedfirms
TOTALASSETSa 554 162932 21882 571709
157
7429300
SALESa
554 164911 25277 554670
343
7955700
ROAb
554
1572 1526
812 −3308
6069
ROSb
554
1328 1252
830 −6044
5293
Tobin’sQc
501
151
104
149
010
1425
COGS/SALESb
554
6303 6545
1572
1757
9480
SALES/ASSETSb 554
12740 12162
4542
1834
39547
ControlfirmsmatchedbyindustryROAandASSETS
TOTALASSETSa 554 169009 20072 611332
221
8634300
SALESa
554 177013 25416 694121
224 11049600
ROAb
554
1554 1515
791 −3138
5816
ROSb
554
1386 1327
960−11514
5368
Tobin’sQc
515
149
128
098
−083
562
COGS/SALESb
554
6346 6306
1371
1293
14548
SALES/ASSETSb 554
12576 12131
4449
2099
38222
ControlfirmsmatchedbyindustryROA
TOTALASSETSa 554 171334 28953 451950
783
5067463
SALESa
554 172015 33071 475046
536
5318035
ROAb
554
1566 1535
802 −3275
6049
ROSb
554
1269 1304
872−13612
3272
Tobin’sQc
548
157
135
098
011
790
COGS/SALESb
554
6237 6193
1003
3252
9162
SALES/ASSETSb 554
13759 13569
2833
4750
25052
ControlfirmsmatchedbyindustryASSETS
TOTALASSETSa 554 167987 21217 580656
216
6598554
SALESa
554 165642 25487 574450
264
6824178
ROAb
554
1058 1140
701 −4809
2407
ROSb
554
133
973
3381−30228
2622
Tobin’sQc
554
143
131
073
012
482
COGS/SALESb
554
7065 6716
2721
3576
33528
SALES/ASSETSb 554
11782 12169
2546
5511
24246
aTOTALASSETSandSALESareinmillionsofdollars.
bROA,ROS,COGS/SALES,andSALES/ASSETSareexpressedinpercent-
ages.
cTobin’sQisadimensionlessmeasure.
firm’sshort-termliabilitiesnetofitsshort-termassets
plus the book value e oflong-term debt, and TA the
bookvalue oftotal assetsofthe firm. Adams(1999)
usesthissameformulainhisNewZealandstudyof
adoptionofISO9000,andKingandLenox(2002)use
the sameintheirstudyoffinancialeffectsofvarious
typesofpollutionreduction.
5. MethodologyandResults
The idealapproachtotest ourhypotheseswould be
to formulate a panel data model (see, for instance,
Hsiao1993),explainingROAintermsofcertification
and several control variables. A major challenge is
thatthecertificationvariableisdiscreteand endoge-
nous. A commonapproachto drawingcausalinfer-
encesfromnonexperimentaldataisHeckman’s(1978)
two-stagemethod,whichfirstpredictstheprobability
thatafirmwillseekISO9000certification,basedon
VB.NET Image: Image Cropping SDK to Cut Out Image, Picture and
SDK; VB.NET image cropping method to crop picture / photo; VB.NET image cropping control add-on needs a PC com is professional provider of document, content and
how to add an image to a pdf in preview; add a jpeg to a pdf
VB.NET Image: Image Scaling SDK to Scale Picture / Photo
this VB.NET image scaling control add-on, we API, developer can only scale one image / picture / photo at com is professional provider of document, content and
add image in pdf using java; acrobat add image to pdf
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
ManagementScience51(7),pp.1046–1059,©2005INFORMS
1051
itscharacteristics, and thenusesthe predictedprob-
abilityasindependent variable inthe second stage.
Inourcase,thismethodhasseveraldrawbacks.First,
thereexistsnosinglegenerallyacceptedextensionof
thisapproachtoapaneldatacontext,andtheresults
weobtainedwerehighlysensitivetotheexactmodel
specificationemployed.Second,withallspecifications
wetried,thepredictedcertificationprobabilitiesinthe
firststagewereverypoorpredictorsofactualcertifi-
cation, whichfurtherreduced ourconfidence inthis
approachwiththisparticulardataset.Afullanalysis
ofthispaneldatasetwouldrequiremaximumlikeli-
hood estimationwithseveral differentdistributional
assumptionstoexaminesensitivityoftheresults;this
isleftforfutureresearch.
Forthesereasons, we revert to event-studymeth-
ods. We largely follow the approach for detecting
abnormal operatingperformance outlined in Barber
and Lyon (1996). An event study requires specify-
ingtheeventperiod—inourcase,theperiodduring
whichISO9000isimplemented.Atypicalimplemen-
tationtakesanywhere between6and 18months, so
wedefine ourevent period astheyearinwhichthe
certificationwasreceived(yeart)andtheyearimme-
diatelyprecedingcertification(yeart−1).Theperiod
preceding the event period (year t−2) is used for
determining the control group. The three years fol-
lowingtheeventperiod(yearst+1throught+3)are
usedtotestforabnormalperformance.
Barber and Lyon(1996) discuss three key choices
in designing an event study. First, the selection of
ameasure ofoperating performance: they find that
operating income divided byassets, or ROA is the
preferred measure (compared to other similar mea-
suresofprofitability).WereportresultsonROAand
ROS. The second decisionistospecifya firm’spre-
dictedperformance inthe absence ofthe event.This
is generally done by comparing the certified firms’
postevent performance with acontrol group ofone
ormorefirms,normallydrawnfromthesameindus-
try(two-digitSICcode),that didnot experience the
event. BarberandLyon(1996)showthat selectinga
controlgroupwithsimilarpre-eventperformance to
the certified firm drasticallyincreases the power of
thestudy,especiallywhenthecertifiedfirmsalready
havehigherpre-eventperformance,asisthecasewith
ISO9000.Lie(2001)confirmsthatmatchingbyROA
isappropriateinthiscase.
AlthoughBarberandLyon(1996)findthat match-
ing by size in addition to industry and pre-event
performance provides little or no extra benefit, we
report this matchingas our mainresults, giventhat
ourCompustatsampledoesincludesmallfirmslisted
on NASDAQ as well as larger firms listed on the
NYSE and other exchanges. The e control firm is the
firminthesameindustrythatis“closest”tothecer-
tified firm in terms of assets and ROA. We trans-
formedassetsandROAineachyearintoz-scoresby
subtracting the industry mean and dividing by the
industry standard deviation for that year; “closest”
wasthendefinedbythe smallest Euclideandistance
in the two-dimensional space e of z-scores, subject to
the constraint thatthe controlfirm’sROA hadtolie
between 90% and 110% of the e certified firm’sROA,
and its assetshad to lie between 50% and 200% of
the certified firm’s assets. Barber and Lyon (1996)
usedthe samerangeforROA,butonlya70%–130%
range for assets;we expandedtherange inresponse
toareviewer’ssuggestion,asthisreducesthenumber
ofcertified firmsfor which no matchcanbe found.
Expanding the matching range further, for instance
to a factor three from the e certified firm’s assets, as
Hendricks and Singhal (1997) do, yielded veryfew
additionalobservationsinourcase.Manyotherstud-
ies, including Hendricks and Singhal (1997), select
control firms based on industryand size e alone; we
report those resultstoo, aswellas resultsbased on
matching by industry and performance alone. The
controlgroupcanconsistofasinglecloselymatching
firm,asinHendricksandSinghal(1997),orofaport-
folioofcloselymatchingfirms,asinBarberandLyon
(1996); we e report results usingboth methods. Con-
trolportfoliosconsistofallfirmsthatmeettheabove
matchingcriteria,ratherthanjusttheclosestmatch.
There is no unique best wayto choose a control
sample. Another matching method uses propensity
scores, but SmithandTodd(2005)find that the evi-
dencetodateaboutitssuperiorperformanceisques-
tionable. The control firms should be as similar as
possibletothefirmsreceivingISO9000certification,
somatchingonmultipledimensionsispreferred;this
iswhythematchingbyindustry,ROA,andASSETS
providesourmainresults.
Expected performance EP
il
 of firm i in any
period +l, usingperiod  asbase period, isgiven
byEP
il
=P
i
+PI
i+l
−PI
i
,where PI
i
isthe
performance of firm i’s control group in period .
Abnormal performance is then given by AP
il
=
P
i+l
−EP
il
,andtheeventstudyconsistsoftesting
whether this abnormal performance is significantly
differentfromzerointhehypothesizeddirection.
Barber and Lyon’s (1996) third findingis that, in
their study, nonparametric tests are more powerful
thanparametric(t-)tests. Thechoice ofateststatis-
tic depends on the distribution of abnormal per-
formance; in our case, the Shapiro-Wilktest always
rejected the hypothesis that abnormal performance
was normallydistributed with p-values below 0.01,
indicating that nonparametric tests are called for.
When the e distributionis symmetric we can use the
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
1052
ManagementScience51(7),pp.1046–1059,©2005INFORMS
Wilcoxon signed-ranktest; inother cases, the distri-
butionishighlyskewedsoweusethelesspowerful
signtest(Conover1999).Forcompleteness,wereport
allthreestatistics.
We can now test the hypotheses by examining
whether certified firms experience abnormal perfor-
mance in terms of ROA, ROS, Tobin’s Q, COGS/
SALES, SALES, and SALES/ASSETS, respectively.
Tables 2–7 show the results for ROA using all six
matching criteria. Tables 8–12 show the results for
the remaining performance measures, using one-to-
one matching byindustry, ROA, and size. Table 13
summarizestheresultsforROAacrossallsixmatch-
ingmethods. The fullresults are e available in asep-
aratedocument.Foreachtest,wereport thesample
size under “N.” The next two columns report the
Table2
AbnormalPerformanceinROA;One-to-OneMatching,by
Industry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
Na meanb medianb skc (t-test)d (WSRtest)d (signtest)d
t−3to
t−2
405
0742
0423
00695
00595
00745
t−2to
t−1
554
0620
0124
00824
00766
02220
t−1
tot
554
1508
0584
s 00075∗∗
00017∗∗∗
00167∗∗
tto
t+1
363 −0288 −0246
06913
07809
07847
t+1to
t+2
233
0417
0469
02219
01151
00839
t+2to
t+3
147 −0525
0108
07594
06171
04345
t−2
tot
554
2128
0996
00009∗∗∗ 00002∗∗∗
00097∗∗∗
t−2to
t+1
363
1911
1035
00014∗∗
00006∗∗∗
00090∗∗∗
t−2to
t+2
233
3183
2437
00001∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+3
147
3936
2575
s 00001∗∗∗ 00001∗∗∗
00008∗∗∗
t−1to
t+1
363
1343
0953
00177∗∗
00296∗∗
00090∗∗∗
t−1to
t+2
233
2230
2597
00040∗∗∗ 00001∗∗∗
00000∗∗∗
t−1to
t+3
147
3575
3371
s 00010∗∗∗ 00002∗∗∗
00158∗∗
p<
01;∗∗p<005;∗∗∗p<001.In allcases,thehypothesisofnor-
mallydistributedabnormalperformancewasrejectedbytheShapiro-Wilk
testwithp-valuesbelow1%,sothet-testisneverappropriate,butreported
forcompleteness;iftheskewnessissubstantialthesigntestisappropriate,
otherwiseonecanusetheWSRtest.
a“N”indicatestotalnumberofcertifiedfirmsinthecorrespondingsample.
bAP
mean and median indicate the mean and median abnormal
performance.
cAn“s”incolumn“sk”indicatesthattheabsolutevalueoftheskewness
isgreaterthanone.
dThep-valuesshownarethosefortheone-sidedtestofthenullhypothesis
ofnoabnormalperformance,usingat-test,Wilcoxonsignedranktest,and
signtestrespectively.
Table3
AbnormalPerformanceinROA;PortfolioMatching,by
Industry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test) (WSRtest) (signtest)
t−3to
t−2
405 −00173 0.0000
05161
04905
05000
t−2to
t−1
554
08906 0.9124
00102∗∗
00001∗∗∗ 00002∗∗∗
t−1
tot
554
16709 1.3656 s 00018∗∗∗ 00000∗∗∗ 00000∗∗∗
tto
t+1
363
02498 0.3803
03197
00990
00944
t+1to
t+2
233
05664 0.3695
00813
00738
01624
t+2to
t+3
147
00851 0.6468
04519
03445
01241
t−2
tot
554
25615 1.6100
00000∗∗∗ 00000∗∗∗ 00000∗∗∗
t−2to
t+1
363
28874 1.8706
00000∗∗∗ 00000∗∗∗ 00000∗∗∗
t−2to
t+2
233
37216 1.9982
00000∗∗∗ 00000∗∗∗ 00000∗∗∗
t−2to
t+3
147
41047 2.9681 s 00000∗∗∗ 00000∗∗∗ 00000∗∗∗
t−1to
t+1
363
23241 1.6970 s 00000∗∗∗ 00000∗∗∗ 00000∗∗
t−1to
t+2
233
29311 2.1526
00000∗∗∗ 00000∗∗∗ 00000∗∗∗
t−1to
t+3
147
35600 2.9978 s 00003∗∗∗ 00001∗∗∗ 00103∗∗
Table4
AbnormalPerformanceinROA;One-to-OneMatching,by
Industry,ROA
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test) (WSRtest) (signtest)
t−3to
t−2
405 −0926 −0772
09359
09698
09104
t−2to
t−1
554
2666
1301
s 00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1
tot
554
2197
0905
00005∗∗∗ 00001∗∗∗
00042∗∗∗
tto
t+1
363
0978
0028
s 00884
02114
03964
t+1to
t+2
233 −0585
0171
s 07070
08539
03230
t+2to
t+3
147
1701
0444
00750
00619
00346∗∗
t−2
tot
554
4863
2620
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+1
363
5633
2824
s 00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+2
233
5714
2688
s 00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+3
147
8976
3751
s 00000∗∗∗ 00000∗∗∗
00040∗∗∗
t−1to
t+1
363
3424
1364
00004∗∗∗ 00002∗∗∗
00118∗∗
t−1to
t+2
233
3166
2172
00123∗∗
00013∗∗∗
00035∗∗∗
t−1to
t+3
147
6965
4604
00006∗∗∗ 00001∗∗∗
00065∗∗∗
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
ManagementScience51(7),pp.1046–1059,©2005INFORMS
1053
Table5
AbnormalPerformanceinROA;PortfolioMatching,by
Industry,ROA
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test)
(WSRtest) (signtest)
t−3to
t−2
405 −1338 −1304
09998
10000
10000
t−2to
t−1
554
2314
2377
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1
tot
554
1395
1710 s 00004∗∗∗ 00000∗∗∗
00000∗∗∗
tto
t+1
363
0552
0573 s 01303
00219∗∗
00293∗∗
t+1to
t+2
233
0936
0945
00108∗∗
00035∗∗∗
00179∗∗
t+2to
t+3
147
0888
1179
00328∗∗
00048∗∗∗
00008∗∗∗
t−2
tot
554
3709
3550 s 00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+1
363
4577
3934
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+2
233
5760
4458
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+3
147
7039
5037
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1to
t+1
363
2438
2543
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1to
t+2
233
3632
3504
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1to
t+3
147
5280
4353
00000∗∗∗ 00000∗∗∗
00000∗∗∗
Table6
AbnormalPerformanceinROA;One-to-OneMatching,by
Industry,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test)
(WSRtest) (signtest)
t−3to
t−2
405
0530
0559
01542
01026
00896
t−2to
t−1
554
0232
0073
02997
04058
03353
t−1
tot
554
0561
0097
01710
04231
04325
tto
t+1
363
1029
0315 s 00866
02375
03181
t+1to
t+2
233 −0709
0384
08817
04644
02351
t+2to
t+3
147
2766
0800 s 00025∗∗∗ 00112∗∗
00103∗∗
t−2
tot
554
0794 −0151
01307
03626
07380
t−2to
t+1
363
2026 −0533 s 00204∗∗
02535
07847
t−2to
t+2
233
1847
2300
00255∗∗
00208∗∗
00654
t−2to
t+3
147
5825
3610 s 00001∗∗∗ 00002∗∗∗
00005∗∗∗
t−1to
t+1
363
1850
0150 s 00159∗∗
02526
03565
t−1to
t+2
233
1384
1097
00578
00407∗∗
00283∗∗
t−1to
t+3
147
5021
2264 s 00005∗∗∗ 00026∗∗∗
00158∗∗
Table7
AbnormalPerformanceinROA;PortfolioMatching,by
Industry,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test) (WSRtest) (signtest)
t−3to
t−2
405 08539 09685
00103∗∗
00021∗∗∗
00019∗∗∗
t−2to
t−1
554 05841 06719
00385∗∗
00006∗∗∗
00029∗∗∗
t−1
tot
554 02966 04386
02204
00148∗∗
00804
tto
t+1
363 01412 01225
s 03735
03820
03373
t+1to
t+2
233 03289 06346
01481
00295∗∗
00578
t+2to
t+3
147 06515 10701
00727
00204∗∗
00103∗∗
t−2
tot
554 08807 11045
s 00251∗∗
00001∗∗∗
00009∗∗∗
t−2to
t+1
363 07244 03369
00605
00353∗∗
01241
t−2to
t+2
233 13506 11356
00047∗∗∗ 00035∗∗∗
00179∗∗
t−2to
t+3
147 23401 11860
00005∗∗∗ 00018∗∗∗
00494∗∗
t−1to
t+1
363 05884 04219
00812
00906
01469
t−1to
t+2
233 10446 09561
00129∗∗
00039∗∗∗
00019∗∗∗
t−1to
t+3
147 21729 22921
00007∗∗∗ 00007∗∗∗
00040∗∗∗
Table8
AbnormalPerformanceinROS;One-to-OneMatching,by
Industry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test) (WSRtest) (signtest)
t−3to
t−2
405
0266
0024
02340
03584
04603
t−2to
t−1
554
0442
0347
s 02113
00097∗∗∗
00306∗∗
t−1
tot
554
1074
0255
00053∗∗∗ 00030∗∗∗
00252∗∗
tto
t+1
363
0201 −0058
03199
04565
06036
t+1to
t+2
233
0887
0604
s 00579∗∗
00910
00208∗∗
t+2to
t+3
147 −0177
0245
06121
03175
03396
t−2
tot
554
1516
0973
s 00034∗∗∗ 00001∗∗∗
00022∗∗∗
t−2to
t+1
363
2024
0810
00003∗∗∗ 00001∗∗∗
00155∗∗
t−2to
t+2
233
3343
1792
00000∗∗∗ 00000∗∗∗
00004∗∗∗
t−2to
t+3
147
4121
2788
00000∗∗∗ 00000∗∗∗
00000∗∗∗
t−1to
t+1
363
1161
0498
00114∗∗
00081∗∗∗
00201∗∗
t−1to
t+2
233
2161
1682
00026∗∗∗ 00002∗∗∗
00004∗∗∗
t−1to
t+3
147
3275
2069
00002∗∗∗ 00000∗∗∗
00000∗∗∗
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
1054
ManagementScience51(7),pp.1046–1059,©2005INFORMS
Table9
AbnormalPerformanceinTobin’sQ;One-to-OneMatching,by
Industry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test)
(WSRtest) (signtest)
t−3to
t−2
282
0031 −0002
03321
07853
05711
t−2to
t−1
406
0083
0052
00973
00211∗∗
00368∗∗
t−1
tot
468
0097
0065
00315∗∗
00089∗∗∗
00080∗∗∗
tto
t+1
315 −0020
0018
06295
04997
01987
t+1to
t+2
200 −0076 −0003
08763
04930
07146
t+2to
t+3
126 −0122 −0067
09792
09606
09634
t−2
tot
405
0186
0179
00069∗∗∗ 00002∗∗∗
00000∗∗∗
t−2to
t+1
256
0167
0209
00248∗∗
00011∗∗∗
00003∗∗∗
t−2to
t+2
160
0212
0200
00375∗∗
00015∗∗∗
00034∗∗∗
t−2to
t+3
102
0195
0179
00436∗∗
00308∗∗
00364∗∗
t−1to
t+1
298
0057
0057
02176
01279
00656
t−1to
t+2
182
0061
0090 s 02870
00208∗∗
00265∗∗
t−1to
t+3
116
0149
0164
00738
00221∗∗
00124∗∗
Table10
AbnormalPerformanceinCOGS/SALES;One-to-One
Matching,byIndustry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test)
(WSRtest) (signtest)
t−3to
t−2
405 −0432 −0217 s 01155
01140
00970
t−2to
t−1
554 −0215 −0398 s 02566
00191∗∗
00067∗∗∗
t−1
tot
554 −0812 −0242 s 00055∗∗∗ 00253∗∗
00625
tto
t+1
363 −0240 −0299
02011
01052
00413∗∗
t+1to
t+2
233 −0848 −0404 s 00367∗∗
01380
00654
t+2to
t+3
147
0349
0444
07561
07076
08928
t−2
tot
554 −1027 −0533
00037∗∗∗ 00088∗∗∗
00107∗∗
t−2to
t+1
363 −1328 −1143
00024∗∗∗ 00006∗∗∗
00067∗∗∗
t−2to
t+2
233 −2503 −1559
00002∗∗∗ 00000∗∗∗
00000∗∗∗
t−2to
t+3
147 −3283 −2129
00000∗∗∗ 00000∗∗∗
00001∗∗∗
t−1to
t+1
363 −0890 −0592
00135∗∗
00077∗∗∗
00779
t−1to
t+2
233 −1803 −1163
00026∗∗∗ 00010∗∗∗
00006∗∗∗
t−1to
t+3
147 −2348 −1460
00014∗∗∗ 00006∗∗∗
00065∗∗∗
Table11
AbnormalRelativeSALESGrowth;One-to-OneMatching,by
Industry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test) (WSRtest) (signtest)
t−3to
t−2
405 −0023 −0011
08829
09858
08847
t−2to
t−1
554
0010 −0018
02976
08068
09427
t−1
tot
554
0037
0014
00049∗∗∗ 00119∗∗
00625
tto
t+1
363
0049
0015
00071∗∗∗ 00082∗∗∗
00460∗∗
t+1to
t+2
233
0006
0019
04056
01911
01785
t+2to
t+3
147
0044
0034
00225∗∗
00347∗∗
00673
t−2
tot
554
0070
0000
00303∗∗
02785
05000
t−2to
t+1
363
0091
0031
00965
00301∗∗
01342
t−2to
t+2
233
0118
0016
01038
01205
03466
t−2to
t+3
147
0220
0049
s 01099
01585
02533
t−1to
t+1
363
0090
0028
00099∗∗∗ 00018∗∗∗
00938
t−1to
t+2
233
0116
0049
00291∗∗
00280∗∗
01462
t−1to
t+3
147
0236
0116
s 00192∗∗
00253∗∗
00919
Notes.“Abnormalrelativesalesgrowth”isdefinedbycomparingSALES
t
SALES
t−1
/SALES
t−1
ofcertifiedfirmswiththatofthecontrolfirms.
mean and median abnormal performance. The first
rowreports abnormalperformance e (AP)in the e year
priortothematching(uptotwoyearsbeforecertifica-
tion),i.e.,AP
il
with=t−3andl=1.(Wediscuss
laterwhythiswasincluded.)Thesecondrowreports
abnormal performance in the e year immediatelyfol-
lowing the decision to embark on the certification
process—i.e., AP
il
with =t−2 and l=1—the
year before the certification is awarded. The third
row measures abnormalperformance in the year of
certification, i.e., AP
il
with  =t−1 and l=1,
etc. The second and third block of results report
longer-termabnormalperformance; for instance, the
row “−2 to 0” measuresAP
il
with =t−2 and
l=2. The column marked “sk”showsan “s” if the
absolute skewness of the distribution of abnormal
performance, defined as the third moment of the
distributiondividedbythecubeofthestandarddevi-
ation, is greater than one; Hair et al. (1998) sug-
gest that indicates severe skewness. The next three
columns report the p-values corresponding to para-
metricandnonparametrictestsofthehypothesisthat
certifiedfirmsexhibitsuperiorperformance(i.e.,pos-
itiveabnormalperformanceforROA,ROS,Tobin’sQ,
SALES, and SALES/ASSETSand negativeabnormal
performanceforCOGS/SALES).
Corbettetal.: TheFinancialImpactofISO9000CertificationintheUnitedStates
ManagementScience51(7),pp.1046–1059,©2005INFORMS
1055
Table12
AbnormalPerformanceinSALES/ASSETS;One-to-One
Matching,byIndustry,ROA,ASSETS
From
AP
AP
p-value
p-value
p-value
year
N
mean
median sk
(t-test)
(WSRtest) (signtest)
t−3to
t−2
405
2468
0891
00765
01579
01850
t−2to
t−1
554 −2511 −0947
09648
08630
08463
t−1
tot
554
3486
3472
00039∗∗∗ 00004∗∗∗
00008∗∗∗
tto
t+1
363 −1506 −0863
08252
08400
09221
t+1to
t+2
233 −3674 −0555
09672
07120
06770
t+2to
t+3
147 −4926 −2601
09956
09943
09717
t−2
tot
554
0975
1097
02636
01619
01344
t−2to
t+1
363 −0500
0250
05935
04675
03369
t−2to
t+2
233 −2465
0004
07836
04360
04738
t−2to
t+3
147 −6796 −3309 s 09436
08368
07187
t−1to
t+1
363
2152
2156
01256
01042
00413∗∗
t−1to
t+2
233 −1198
2428
06666
02602
03230
t−1to
t+3
147 −1833 −2469
06967
04527
07187
6. DiscussionandLimitations
6.1. DiscussionofResults
In this section, we discuss our main results, after
whichweaddsomenotesaboutinterpretation.Table2
Table13
SignificanceofAbnormalPerformanceinROAUnderAllSixMatchingCriteria
Industry,ROA,ASSETS
Industry,ROA
Industry,ASSETS
One-to-one
Portfolio
One-to-one
Portfolio
One-to-one
Portfolio
Fromyear
Na
WSRb signb WSRb
signb
WSRb signb WSRb signb WSRb
signb WSRb signb
t−3tot−2
405
∗∗∗
∗∗∗
t−2tot−1
554
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
t−1tot
554
∗∗∗
∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
ttot+1
363
∗∗
∗∗
t+1tot+2
233
∗∗∗
∗∗
∗∗
t+2tot+3
147
∗∗
∗∗∗
∗∗∗
∗∗
∗∗
∗∗
∗∗
t−2tot
554
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
t−2tot+1
363
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
t−2tot+2
233
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
∗∗∗
∗∗
t−2tot+3
147
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
t−1tot+1
363
∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
∗∗∗
∗∗∗
t−1tot+2
233
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
∗∗
∗∗∗
∗∗∗
t−1tot+3
147
∗∗∗
∗∗
∗∗∗
∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗
∗∗∗
∗∗∗
p<01;∗∗p<005;∗∗∗p<001.
a“N”indicatestotalnumberofcertifiedfirmsinthecorrespondingsample.
bThesignificancelevelsshownarethosefortheone-sidedtestofthenullhypothesisofnoabnormalperformance,usingthe
Wilcoxonsignedranktestandsigntest,respectively.
showsthatwhenselectingsinglecontrolfirmsbased
on industry, ROA, and size, firms experienced sig-
nificant abnormalimprovementsinROAinthe year
of their first ISO 9000 certification. Using the sign
test, the most conservative of the three, abnormal
performance for t−1 to t, the year leading up to
certification, isweaklysignificant p=00167. After
certification, no further abnormal improvements in
ROAoccur.Thecumulativeimprovements,however,
appear to be strong and lasting, as all the longer-
termtests,usingeithert−2or t−1asbaseperiod,
show significant abnormal improvements in ROA,
usuallywithp-valuesbelow 0.01. The magnitude of
the longer-term effects is considerable, in the e order
of 1–4 percentage points. Starting with an ROA of
15.7%(themeanofcertifiedfirmsinyeart−2,from
Table1a),themeanabnormalincreaseinROAbyyear
t+3is3.94percentagepoints(Table2),comparedto
noncertifiedfirmsstartingwiththesameROA,which
correspondstoarelativeincreaseof25%.
Tables 3–5 show that matching by industry and
ROAalone,and/orusingportfoliosofcontrolfirms,
haslittleeffectonthelong-termresults,althoughthe
pureperformance-matchedsampleexhibitsalevelof
abnormal performance that isalmost twice ashigh.
The short-termeffects do become much stronger in
allthree cases, inthe year ofcertification(t−1to t)
and now also inthe precedingyear (t−2 to t−1).
Matching by size alone using portfolios of control
firmsalsoleadstosimilarresults,asshowninTable7.
However, one-to-one matching based on size alone
(Table6)makestheshort-termeffectsdisappear;even
there, though, the long-term effects remain strong.
We discussthedifferencebetweenthemethodslater.
Documents you may be interested
Documents you may be interested