the sorenson workstation five data

TheSorensonWorkstationFIVEDataby

OlavSorenson

JeffreyS.SkollChairinTechnicalInnovationandEntrepreneurshipProfessorofStrategicManagement

JosephL.RotmanSchoolofManagementUniversityofToronto

©2008

SorensonWorkstationFIVEData2

Copyright©2008OlavSorenson FIVEProjectDocumentation

Contents 1. Conditions of Use................................................................................................ 32. Data Set Author ................................................................................................... 43. Data Set Summary............................................................................................... 44. Author’s Research Using This Data Set.............................................................. 45. Additional References ......................................................................................... 56. Data Set Sources.................................................................................................. 57. Data Description.................................................................................................. 5

7.1. Product/Industry Definition........................................................................... 67.2. Excluded Fields ............................................................................................. 7

8. Tables .................................................................................................................. 88.1. Descriptive Statistics, File 1: SorensonworkprodFIVEdata......................... 88.2. Descriptive Statistics, File 2: SorensonworkfirmFIVEdata ......................... 8

9. Variable List and Definitions: File 1: SorensonworkprodFIVEdata.................. 99.1. Year of Observation: ..................................................................................... 99.2. Firm Identifiers:............................................................................................. 99.3. Industry Submarket or Product Identifiers: ................................................. 109.4. Other Variables: .......................................................................................... 10

10. Variable List and Definitions: File 2: SorensonworkfirmFIVEdata .............. 1410.1. Year of Observation: ................................................................................. 1410.2. Firm Identifiers:......................................................................................... 1410.3. Entry/Exit Variables: ................................................................................. 1510.4. Firm Size Variables: .................................................................................. 1610.5. Location Variables: ................................................................................... 1610.6. Research and Development Variables:...................................................... 1810.7. Products and Markets Variables:............................................................... 1810.8. Other Variables: ........................................................................................ 18



1.ConditionsofUseThedatasetsdescribedinthisdocumentarepartoftheFIVEProjectonFirmandIndustry Evolution and Entrepreneurship, organized by Constance Helfat (TuckSchool of Business at Dartmouth) and Steven Klepper (Carnegie‐MellonUniversity).Ifyoudownloadorotherwiseobtainanyofthedatasetsdescribedin thisdocument,orobtainmodifiedversionsof thesedata sets, youagree toabidebytheconditionsofusesetforthhere.TheorganizersoftheFIVEProject,ConstanceHelfatandStevenKlepper,arenotresponsibleforthecontentofanyoftheFIVEProjectdatasets.TheFIVEProjectorganizers and the FIVE Project data set authors make no representations orwarranties of any kind concerning anyof the FIVE Project data, includingwithregardtotheabsenceorpresenceoferrors.AllFIVEProjectdatasetsareprotectedbycopyright.Thecopyrightholders foreachdatasetareindicatedonthefirstpageofthedocumentationforeachdataset. Permission is granted to reproduce FIVE Project data for non‐profiteducationalandresearchuseonly.RequeststoreproduceFIVEProjectdataforotheruses shouldbeaddressed to:ProfessorConstanceHelfat,TuckSchoolofBusinessatDartmouth,Hanover,NH03755,constance.helfat@dartmouth.edu.ThisdocumentdescribestheSorensonWorkstationFIVEData,whichshouldbereferred toby thisname inanyderivativeworks. Inanywrittenandpublishedwork,usersofthesedatashouldcitethisdocument,oneofthedatasetauthor’soriginalarticles(eitherSorenson,2000,orSorenson,2003),andtheFIVEProject:DataOverview(Helfat&Klepper,2007).Youmaycreateanewdata setbymodifyingaFIVEProjectdata set, includingbutnotlimitedtoaddingorremovingdataormergingdatafromdifferentFIVEProject data sets. If you utilize any FIVE Project data to create new data sets,withinsixyears fromthedate thatyouobtain theFIVEProjectdata,youmustprovide the FIVE Projectwith a copy of yourmodified data set by contacting:ConstanceHelfat,[email protected] written documents or statistical estimates that use FIVE Project data,includingbutnot limitedtoworkingpapersandpublications,mustbereportedto:ConstanceHelfat,[email protected](specifiedbytheauthorofthisdocument):IfyousupplementtheSorensonWorkstationFIVEData,youmust renamethatvariantof theoriginaldatawithSorensonappearing first in the listofdatasetauthors, alongwith thewordFIVE. Forexample, if JoeCampusaddedproduct



performancecharacteristics,thedatasetwouldbecalledthe“Sorenson‐CampusWorkstationFIVEData.”IfyoumergetwoormoreFIVEdatasets,thenewnameofthedatasetmustincludethenamesofalloftheoriginaldatasetauthorsandthewordFIVE.

2.DataSetAuthorOlavSorensonJeffreyS.SkollChairinTechnicalInnovationandEntrepreneurshipProfessorofStrategicManagementJosephL.RotmanSchoolofManagementUniversityofToronto

3.DataSetSummaryThedatadescribeallfirmsthatIcouldidentifysellinga“computerworkstation”at any point in time between 1980, the genesis of the industry (for a briefdescription of the history of the industry, see Sorenson, 2003), and 1996. Intotal,itcovers175firmsfor690firm‐years(DataFile2)and1,276productsfor2,735 product‐years (Data File 1). SeeDataDescription for a discussion of thedefinitionofa“computerworkstation”andanexplanationoftheprocessusedtoidentifycomputerworkstationsandthefirmsthatproducedthem.

4.Author’sResearchUsingThisDataSetSorensonO.2000.Lettingthemarketworkforyou:anevolutionaryperspective

onproductstrategy.StrategicManagementJournal21:577‐592.SorensonO.2003.Interdependenceandadaptability:organizationallearning

andthelong‐termeffectofintegration.ManagementScience49:446‐463.

SorensonO.2005."Socialnetworksandthepersistenceofclusters:Evidencefromthecomputerworkstationindustry,’pp.297‐316inClusters,networksandinnovation,BreschiS,MalerbaF(eds).OxfordUniversityPress:NewYork.

SorensonO,McEvilyS,RenCR&RoyR.2006.Nichewidthrevisited:organizationalscope,behaviorandperformance.StrategicManagementJournal27:915‐936.



5.AdditionalReferences

BarronDN,WestE&HannanMT.1994.Atimetogrowandatimetodie:GrowthandmortalityofcreditunionsinNewYorkCity,1914‐1990.AmericanJournalofSociology100:381‐421.

BersteinResearch.1992.TheWorkstationMarket.SanfordC.Berstein&Co.:NewYork.

Helfat,CE&Klepper,S.2007.FIVEProject:DataOverview.http://papers.ssrn.com/paper=1028022.

SunMicrosystems.1986.Sunnews(Tech.Rep.).SunMicrosystems:PaloAlto,CA.

6.DataSetSourcesTheprimarysourceforidentifyingproductsisDataSources,acatalogpublishedby Information Access Company of Foster, CA at varying frequencies – eithersemi‐annually or quarterly – over the period covered by the data. I identifiedadditional products within this category by reading product reviews andscanningadvertisementsinByteandIEEEGraphicalComputingandApplicationsfor the entire period (Byte did not yield information on any products not alsofound in IEEE). I should note that although these sources yielded additionalproducts from 1980‐1982, they provided little more than supplementalinformation from1983onward. I collectedadditional company leveldata fromcorporate reports and from articles available on Lexis‐Nexis (for each firm, Iperformed a search in Lexis‐Nexis using the company name listed in DataSources).

7.DataDescriptionI originally collected these data to understand better how vertical integrationinfluencestheabilityoforganizationstolearnfromexperience(Sorenson,2003).With respect to this question, I chose the computer workstation industrybecauseproductdirectoriesallowedme todetermine tosomeextentwhetherthe firms in the industry produced the components used in theirmachines orwhethertheypurchasedthemfromothercompanies.Thoughoriginallydesignedtoinvestigateintegrationandlearning,Ihavealsousedthesedatatoexploretheeffects of product line breadth and product innovation on firm performance(Sorenson, 2000, Sorenson et al, 2006) and the evolution of the industrygeographically (Sorenson, 2005). Much of the data nonetheless remainsunexploited. Inparticular,very littlehasbeendonewiththeproduct leveldata



set, and with the exception of vertical integration, most of the firm‐levelvariableshavebeenusedonlyascontrols.

7.1.Product/IndustryDefinitionAlthough computer workstations formed an established product category by1985, even the term “computer workstation” did not exist at the time ofintroduction of what experts generally consider the first product, the ApolloDOMAIN.Somefurnituremanufacturersdid,however,usetheterminthe1970sand early 1980s to refer to desks designed to house terminals. Moreover,despitethecommonusageofthetermcomputerworkstationinthepress,DataSources never adopted it as a product category (the products found hereappeared under either “Desktop Systems” or “Graphics Systems”). Given theabsence of a standard catalog, and to capture the early period of themarket/industrybeforeaconsistentcategorynamehadbeenadopted,Idecidedtocollectinformationoncomputersystemswithaconsistentsetofattributes.Definitionsofcomputerworkstationsabound.Mostofthesedefinitionshavethedisadvantage of being relevant at some particular cross‐section in time andtherefore not particularly useful in collecting longitudinal data. For example,Berstein, 1992, defined the workstation product class by the followingcharacteristics:a32‐bitCPU;stand‐alonesingle‐usercapability;multi‐tasking32‐bit operating system; bit‐mapped graphics; utilization of virtual memory; andbuilt‐in networking. This definition posesmultiple problems. First, Bernsteinrequiresthatworkstationsbe32‐bitmachines,butsomeearlyofferingshad16‐bitprocessorsandarchitectures.Forexample,theOrcatechDesignEngineeringWorkstation, introduced in 1982 and one of the first products with“workstation” in its name, had a proprietary 16‐bit processor. This firmeventuallyexited strategically from theworkstationmarket,but for twoyears,thepressciteditasoneoftheup‐and‐comersthatcouldpotentiallyrivalApollo.At the other end, by 1996 most of the major RISC chip manufacturers hadalreadystartedproducing64‐bitchipsfortheworkstationmarket.Second,thisdefinition requires machines to use bit‐mapped graphics. Although allworkstationsonthemarkettodayusethistechnology,someoftheearlyrivalstoApollochosetodesigntheirsystemsaroundrastergraphics.Theubiquityofbit‐mappedgraphicsthereforeappearsanartifactofhindsightontheevolutionoftheproduct.Ithereforedecidedtodefinetheproductaccordingtosomeabstractfeaturesofits functionality, rather than according to specific components. In particular, Idefined a “computer workstation” as a distributed computing machine forprimary use by a single user. This definition differentiated workstationsimportantlyfromthreeotherclassesofcomputingmachines:terminals,servers,and personal computers. Terminals do not have significant local processingcapabilities. Although in the1980’s terminalsevolved to includehardware for



processing some graphics rendering locally, these machines still relied on aserver for all general processing. Servers meanwhile are used in distributedcomputing,butcomputermanufacturersdonotdesignthemforusebyasingleuser; multiple users access their resources through either terminals orworkstations.Personalcomputers,whereallprocessing,memory,andmostdatastorage occur on the user'smachine, sit at the opposite endof the spectrum.Thoughintendedforasingleuser,theyinvolvenodistributedprocessing.Meetingthisdefinitionrequiresbothspecifichardwareandsystemsoftware.Onthe hardware side, workstations must have a CPU and memory for generalprocessing and a monitor for display. These hardware requirements help todistinguishworkstationsfromterminalsandservers.TerminalsdonothaveCPUsfor general processing and servers often have no built‐in display capabilities.Workstations also need an operating system that allows for distributedcomputing. Experts considered the ability to share virtual memory and fileresources across machines the key attribute of the Apollo system, widelyconsidered the first workstation. UNIX did not even provide a distributedcomputingenvironmentuntilSunMicrosystemsintroducedtheNFSprotocolforfile‐sharing in November of 1984, which other manufacturers offering UNIXsystems quickly adopted (Sun, 1986). Although personal computers often lookverysimilartoworkstationsintheirphysicalconfigurations,MS‐DOS,MacOSandother personal computer operating systems did not allow for this seamlesssharingofresourcesduringthe1980sandearly1990s.Becauseof thedefinitionthat Iadopted,1996servesasanaturalendpoint forthedataset.AfterthearrivaloftheWindowsNToperatingsystem(firstreleasedin 1993, but not widely adopted until version 4.0 became available in 1996),distinctionsbetweenpersonalcomputersandcomputerworkstationsblur—oratleast they shift from being qualitative, in terms of functionality, to beingquantitative, in termsofspeedandprice. Ibelievethat thisblurringaccuratelyreflectsthedemiseoftheproductasadistinctcategory/industry.Iusedthesehardwareandsoftwarefeaturestoidentifycomputerworkstationsattheproductlevel.Thoughtheclassificationprocessoftenrequiredadditionalresearch, I adopted the following baseline approach: (1) From the “DesktopSystems” category, I included all systems with operating systems capable ofdistributedcomputing (exceptMacOSandWindowsNT),butexcludedsystemswith “Server” in the name or without graphics capabilities or designated as amultiple user system. (2) From the “Graphics Systems” category, I included allsystemswithoperatingsystemscapableofdistributedcomputing.

7.2.ExcludedFieldsInthepapersthatIhavewrittenfromthesedata,usersmaynotethatImentionandsometimesuseproduct‐levelunitsales.IobtainedthatinformationfromIDC



andthetermsoftheagreementdonotallowmetograntaccesstoothers.Evenif the intellectualproperty issues couldbe resolved, Ihaveunfortunatelybeenunableto locateanuncorruptedfilecontainingthis information,so Iamafraidthatitmayhavebecomeunrecoverable.

8.Tables

8.1.DescriptiveStatistics,File1:SorensonworkprodFIVEdata

Variable N Mean SD Min MaxClockspeed mhz 956 40.8 42.4 1 300Memory ram 2355 131,178 843,323 64 1.60E+07Storage hd 1521 2,980 13,998 20 138,600Display monitor 2354 0.26 0.44 0 1Listprice price 2613 27,389 31,963 400 600,000

8.2.DescriptiveStatistics,File2:SorensonworkfirmFIVEdata

Variable N Mean SD Min Max

year 690 1990 3.9 1980 1996

found 672 1972 20.2 1880 1996

employ 580 15,800 54,241 1 405,535

sales 630 2.70E+09 .9.9e+09 0 7.50E+10

RandD 105 1.90E+08 9.90E+08 294,877 6.40E+09

products 690 22.5 75 0 598

ptypes 690 3.8 5.3 0 33

system 690 0.41 0.49 0 1

graphic 690 0.85 0.36 0 1

cpu 690 0.02 0.13 0 1

os 690 0.25 0.43 0 1

apps 690 0.59 0.49 0 1

comm 690 0.19 0.39 0 1

monitor 690 0.11 0.32 0 1

disk 690 0.16 0.36 0 1

memory 690 0.08 0.27 0 1

board 690 0.11 0.32 0 1



9.VariableListandDefinitions:File1:SorensonworkprodFIVEdataDescriptivestatisticsforthecontinuousandindicatorvariablesbelowappearintheTablessection.

9.1.YearofObservation:

Year[Variablename:year;4digitnumericvariable]:Yearofobservation,rangingfrom1980to1996.

9.2.FirmIdentifiers:

FirmName[Variablename:firmname;40characterstringvariable]:Nameofcompany.

FIVEFirmID(uniquefirmidentifierfortheFIVEProject)[Variablename; fivefirmID;6digitnumericvariable]:Unique identifier foreachFirmNamethatappearsintheFIVEdatasets.FirmNamesandFIVEFirmIDsmayappearinmorethanoneFIVEdataset.Ifafirm’snamechangedovertime,theFIVEFirm ID for the firmsometimesmaychangeaswell. Inaddition,differentFIVE data set authors may have coded different firm names for the samecompany, which could result in different FIVE Firm IDs for the same firm indifferentdatasets.

DataSetFirmID(firmidentifierassignedintheoriginaldataset)[Variablename:firmID;3digitnumericvariable]:UniqueidentifierthatIcreatedforeachfirmintheoriginaldataset.

CUSIPHeaderID(firmidentifierassignedbyCompustat)[Variablename:CUSIPheaderID; 8digitnumeric variable]:Unique identifier foreach firm assigned by Compustat for 2007. For a given Firm Name, this ID isconstant across all years in the data set. These IDs come frommatching eachFirmName in theFIVEdata sets toan identicalor similarname inCompustat.Thereisnoguaranteethatthesematchesareaccurate.TherealsoaremanyFirmNamesforwhichnomatchwasfound.CUSIPHeaderFirmName(companynameassignedbyCompustat)[Variablename:CUSIPheadername;charactervariable]:NameofcompanygiveninCompustatfortheCUSIPHeaderID.



CUSIPHistoricalID(firmidentifierassignedbyCompustat)[Variablename:CUSIPhistoryID; 8 digit numeric variable]:Unique identifier foreach firm assigned by Compustat in their historical databases. For each FirmName, the CUSIP Historical ID is that shown in Compustat for the year ofobservationintheFIVEdataset.TheseIDscomefrommatchingeachFirmNamein theFIVEdatasets toan identicalor similarname inCompustat.There isnoguarantee that this match is accurate. There also are many Firm Names forwhichnomatchwasfound.CUSIPHistoricalFirmName(companynameassignedbyComputstat)[Variablename:CUSIPhistoryname;charactervariable]:NameofcompanygiveninCompustatfortheCUSIPHistoricalID.

9.3.IndustrySubmarketorProductIdentifiers:

ProductName:[Variable name: productname; 40 character string variable]: Either the namelisted inData Sourcesor thenameappearing in the advertisementor productreviewfromwhichIidentifiedthecomputerworkstation.

ProductID(productidentifierassignedintheoriginaldataset)[Variable name: productID; 4 digit numeric variable]: Unique identifier that Icreated for each product. Note that although these IDs uniquely identifyproducts, theymaynotnestwithin companies; several companies continue tosell existing systems under the same names after acquiring or merging withanothercompany.

9.4.OtherVariables:

OperatingSystem[Variable name: os; 2 digit numeric variable]: Operating system installed onsystembythefirmsellingtheworkstation.Thetablebelowprovidesthecodes,names, year of first appearance, number of product‐yearswith that operatingsystem,andwhethertheoperatingsystemhadUNIXasitsunderlyingengine.



Code OSname Firstyear Number UNIX‐based

1 SunOS 1985 477 Yes2 Solaris 1990 147 Yes4 DECOSF 1994 10 No5 MAX/OS 1990 13 Yes?7 DECUNIX 1986 209 Yes8 ULTRIX 1990 50 Yes9 HP/UX 1984 225 Yes10 AIX 1988 120 Yes11 OtherUNIX 1982 708 Yes12 DGAviion 1986 41 No13 DEC3000 1994 27 No14 MicroVMS 1983 317 No15 SGI 1984 23 Yes16 EP/IX 1991 17 Yes18 Classix 1991 13 Yes19 DOMAIN 1983 116 No20 Other 1980 68 No21 Aegis 1981 33 No22 RainbowTOS 1990 7 No23 CTOS 1983 4 No24 NextStep 1990 5 Yes25 OrcaOS 1983 9 No26 CP/M 1982 3 No

Missing 79

CPU[Variable name: cpu; 2 digit numeric variable]: Central processing unit used inmachine.Thetablebelowprovidesthecodes,names,numberofproductswiththat CPU, whether the chip has a RISC architecture (i.e. “Reduced InstructionSet”),andthedesigner(andusuallymanufacturer)ofthechip.Earlyworkstations incorporatedprocessors thathadbeendeveloped forotherapplications,but the industryquicklybecameadriving force inmicroprocessordevelopment. The first Apollo machine as well as most of the other earlycomputerworkstationsusedtheMotorola68000processor.Thischip,releasedin1979,offeredthefirstmassproduction32‐bitprocessor(Inteldidnotreleasea32‐bitprocessoruntil the80386 in1984).RISC technologydidnotappear intheindustryuntil1988,whenSunMicrosystemsbegantopushRISCtechnology.Itneverthelessbecamethedominantarchitectureby1993.



Code CPUname Firstyear Number Architecture Designer

1 SPARC 1987 322 RISC Sun2 SuperSPARC 1993 125 RISC Sun3 ALPHA 1994 74 RISC DEC4 IBMPower 1994 37 CISC IBM5 microSPARCII 1993 54 RISC Sun6 HyperSPARC 1995 12 RISC Sun8 MIPSR4400 1993 49 RISC MIPS9 MIPSR3000 1990 175 RISC MIPS10 MIPSR4000 1992 42 RISC MIPS11 PA‐RISC 1993 53 RISC HP12 PowerPC604 1995 15 RISC IBM13 SuperSPARCII 1996 2 RISC Sun14 Pentium 1995 28 CISC Intel15 Intel80960 1992 143 CISC Intel16 MC68020 1984 328 CISC Motorola17 MC68030 1986 163 CISC Motorola18 MC68040 1991 34 CISC Motorola19 MC88100 1992 67 CISC Motorola20 AlphaAXP 1994 24 CISC DEC21 CVAX 1986 72 CISC DEC22 MC68000 1980 85 CISC Motorola23 SPARC64 1996 2 RISC Sun24 RS6000 1994 10 CISC IBM27 TMS320C65 1991 6 CISC TI28 MIPSR4600 1995 11 RISC MIPS29 MIPSR8000 1994 15 RISC MIPS30 Unknown 1980 402 Unknown Unnamed31 TMS34020 1989 29 CISC TI32 LSIR33020 1994 6 CISC LSI33 80486 1991 21 CISC Intel34 AMD29000 1995 14 CISC AMD35 MIPSR6000 1991 3 RISC MIPS36 80860 1992 8 CISC Intel37 MIPSR2000 1990 14 RISC MIPS38 MC88000 1990 34 CISC Motorola39 MC68010 1983 71 CISC Motorola40 80286 1987 32 CISC Intel41 8085 1991 6 CISC Intel42 ClipperC300 1990 11 CISC Intergraph43 ClipperC400 1993 3 CISC Intergraph44 Sun‐3/4 1990 3 CISC Sun45 NECV30 1988 5 CISC NEC46 AMD2901 1990 14 CISC AMD47 ClipperC100 1990 4 CISC Intergraph48 8086 1983 10 CISC Intel49 Whizzard 1981 47 CISC Megatek50 MicroVax 1983 14 CISC DEC51 Apollo32 1982 10 CISC Apollo52 ApolloS 1986 12 CISC Apollo



Architecture[Variable name:bus; 2 digit numeric variable]: Architecture for integrating thevarious components of the computer system. The most common buses inpersonalcomputerswereISA,thenEISA(andIBM'sproprietaryMCA)andthenPCI,butcomputerworkstationmanufacturerstendedtodesigntheirownbusesto improve system performance. The table below provides the codes, names,yearoffirstappearance,andthenumberofproduct‐yearswiththatbus.

Code Busname Firstyear Number

1 VMEbus 1982 504

2 Sbus 1990 376

3 ISAw/PCI 1985 86

4 Mbus 1993 8

5 ProprietaryVME‐type 1994 5

6 EISA 1992 67

7 Unnamedproprietary 1980 1,318

8 PCIlocalbus 1994 28

9 MCA 1992 97

10 NuBus 1990 511 Multibus 1983 4

Missing 237

ClockSpeed[Variablename:mhz;3digitnumericvariable]:Clockspeedofmachineincycles(megahertz).NotethatonecannotreadilycomparethespeedofCISCandRISCmachinesbecauseoftheirarchitecturaldifferences.

Memory[Variable name: ram; 7 digit numeric variable]: Amount of random accessmemoryinstalledinthestandardconfigurationofthemachine(inkilobytes).

Storage[Variablename:hd;6digitnumericvariable]:Sizeofharddriveinstalledonthemachineinmegabytes.

Display[Variable name:monitor; 1 digit numeric variable]: A code indicatingwhetherthesystemhasanexternalmonitor(=0),oraninternalmonitor(=1).



ListPrice[Variable name: price; 6 digit numeric variable]: List price of standardconfiguration inU.S. dollars. Pleasenote that I havenotdeflatedanyof theseamounts to adjust for inflation. Though not consistent across manufacturers,thesepricesgenerallydonotincludethecostofthemonitor.

10.VariableListandDefinitions:File2:SorensonworkfirmFIVEdataDescriptivestatisticsforthecontinuousandindicatorvariablesbelowappearintheTablessection.

10.1.YearofObservation:

Year[Variablename:year;4digitnumericvariable]:Yearofobservation,rangingfrom1980to1996.

10.2.FirmIdentifiers:

FirmName[Variablename:firmname;40characterstringvariable]:Nameofcompany.

FIVEFirmID(uniquefirmidentifierfortheFIVEproject)[Variablename; fivefirmID;7digitnumericvariable]:Unique identifier foreachFirm Name that appears in the FIVE data sets. Firm names and FIVE firmidentifiersmayappearinmorethanoneFIVEdataset.Ifafirm’snamechangedovertime,theFIVEFirmIDforthefirmsometimeschangesaswell.Inaddition,different FIVE data set authorsmay have coded different FirmNames for thesamecompany,whichcouldresultindifferentFIVEFirmIDsforthesamefirmindifferentdatasets.

DataSetFirmID(firmidentifierassignedintheoriginaldataset)[Variablename:firmID;3digitnumericvariable]:UniqueidentifierthatIcreatedforeachfirmintheoriginaldataset.

CUSIPHeaderID(firmidentifierassignedbyCompustat)[Variablename:CUSIPheaderID;8digitnumericvariable]:UniqueidentifierforafirmassignedbyCompustatfor2007.Foragivenfirmname,thisIDisconstantacrossallyearsinthedataset.TheseIDscomefrommatchingeachFirmNameintheFIVEdatasetstoanidenticalorsimilarfirmnameinCompustat.ThereisnoguaranteethatthesematchesandIDsareaccurate.Ifnomatchwasfound,theCUSIPHeaderIDisblank.



CUSIPHeaderFirmName(companynameassignedbyCompustat)[Variablename:CUSIPheadername;charactervariable]:NameofcompanygiveninCompustatfortheCUSIPHeaderID.

CUSIPHistoricalID(firmidentifierassignedbyCompustat)[Variablename:CUSIPhistoryID;8digitnumericvariable]:UniqueidentifierforafirmassignedbyCompustat in their historical databases. For each FirmName,the CUSIP Historical ID is that shown in Compustat for the year of eachobservationintheFIVEdataset.TheseIDscomefrommatchingeachFirmNameintheFIVEdatasetstoanidenticalorsimilarfirmnameinCompustat.ThereisnoguaranteethatthesematchesandIDsareaccurate.Ifnomatchwasfoundfor a given Firm Name in a given year, the CUSIP Historical ID is blank. Thismeans thata firmmayhaveCUSIPHistorical IDsassigned for somebutnotallyearsinwhichthefirmappearsinthedataset.CUSIPHistoricalFirmName(companynameassignedbyComputstat)[Variablename:CUSIPhistoryname;charactervariable]:NameofcompanygiveninCompustatfortheCUSIPHistoricalID.

10.3.Entry/ExitVariables:

YearofFounding[Variable name: found; 4 digit numeric variable]: Year in which the companybegan operations (not the year that it entered the computer workstationmarket). I assessed market entry from the product‐level data as the year inwhichitfirstofferedacomputerworkstationproduct.For some analyses (Sorenson, 2000), I created either a continuous or dummyvariablefordealiostatus.Thecontinuousversionsimplycalculatedthenumberof years the firm had been in operation prior to entering the computerworkstationmarket.Thedichotomousversioncollapsedthismeasureto0versus1(foranynumberofyearsgreaterthanzero).

NotesonExitEvent[Variablename: fate;50character stringvariable]:Noteson thenatureofexitevents,acquisitions,namechanges,etc.

NatureofExitEvent[Variablename:estate;1digitnumericvariable]:Acodeindicatingwhethertheobservationisleft‐censored(=0),hasexitedthemarket(=1),hasbeenacquiredandcontinuedoperations (=2),hasspunoffandchanged itsname(=3),orhaschangeditsname(=4).Inexitrateanalyses,Igroupedacquisitionswithfailures.



In 13 of the 14 cases, the press discusses the firms' financial troubles prior tobeingacquired.

10.4.FirmSizeVariables:

Revenue[Variablename:sales;10digitnumericvariable]:CompanysalesinU.S.dollars;these figures also include sales outside of the computer workstation market.PleasenotethatIhavenotdeflatedanyoftheseamountstoadjustforinflation.For public companies, these figures came from annual reports. For privatecompanies,theycamefromDataSourcesorLexis‐Nexissearches.Firmswithoutinformationtendtobequitesmall.Todealwiththemissingcasesinanalysesthatusethisvariableasacontrolinmydissertation,Iassignedthemarandomvaluefromthebottomdecileoffirmswithinformationavailableforaparticular timewindow: 351,000 to 2,475,000 formissing fields from 1980 to1984;867,000to3,400,000for1985to1988;180,000to4,900,000for1989to1992;and180,000to5,400,000for1989to1992.Thelowendofthesizerangedropsbeginningin1989duetotheentryof“clone”manufacturers.SeeBarronetal,1994,foradiscussionofthistechniquefordealingwithmissingfields.Testsofthevalidityofthismethodrevealednosignificantdifferencebetweentheexitratesofthesefirmsandthoseofotherfirmsinthebottomdecile(basedonat‐testofadummyvariableforthosecaseswithrandomlyassignedsizes).

NumberofEmployees[Variablename:employ;6digitnumericvariable]:Numberofpeopleemployedatthefirminagivenyear;thesefiguresalso includeemployeesoutsideofthecomputerworkstationdivisionforcompanieswithotherproducts.

10.5.LocationVariables:

ZipCode[Variable name: zip; 5 digit numeric variable]: Zip code of corporateheadquarters.Withtheexceptionofa fewofthe largest firms,almostnoneofthe companies in the data have more than one facility. In general, thoseinterested in thegeographyof the industrywill find the latitudeand longitudemeasuresmoreuseful.



Figure1:Distributionoffirmsin1984

Figure2:Distributionoffirmsin1996

Latitude[Variablename:lat;7digitnumericvariable]:Latitudeoftheheadquartersinradians.

Longitude[Variablename:lon;7digitnumericvariable]:Longitudeoftheheadquartersinradians.



10.6.ResearchandDevelopmentVariables:

R&DExpenditure[Variable name:RandD; 9 digit numeric variable]: Amount ofmoney spent onR&Dat the firm inagivenyear; these figuresalso includespendingoutsideofcomputerworkstations.

10.7.ProductsandMarketsVariables:

NumberofComputerProducts[Variable name:products; 3 digit numeric variable]: A count of the number ofproductsthatthefirmoffersinallcategoriescoveredbyDataSources.

NumberofComputerMarkets[Variable name: ptypes; 2 digit numeric variable]: A count of the number ofproductcategoriesofDataSourcesinwhichthefirmoffersatleastoneproduct.

ProductSection1[Variablename:system;1digitnumericvariable]:Adummyvariable indicatingwhetherthefirmhasproductslistedinthe“DesktopSystems”category.

ProductSection2[Variablename:graphic;1digitnumericvariable]:Adummyvariable indicatingwhetherthefirmhasproductslistedinthe“GraphicsSystems”category.

10.8.OtherVariables:

CEO[Variablename:ceo;30characterstringvariable]:NameofCEO.

Ownership[Variable name:own; 1 digit numeric variable]: A code indicatingwhether thefirmisprivate(=0),public(=1)orasubsidiaryofanotherfirm(=2).

VerticalIntegrationAseriesofindicatorvariablescapturewhetherthefirmproducesproductsthatcouldserveassubsystemsfortheircomputerworkstationproducts.Ingeneral,Idonotknowiftheyincorporatedthesecomponentsorwhethertheyproducedallofthem(e.g.,firmscouldengagein“tapered”integration‐‐‐sourcingaportionoftheirneedsforaniteminternallyandaportionexternally).Thevariablesare:cpu‐centralprocessingunit;os‐operatingsystem;apps‐softwareapplications;comm ‐ communications hardware;monitor ‐monitor;disk ‐ hard disk drives;memory‐randomaccessmemory;andboard‐motherboard.

the sorenson workstation five data

Documents