Metropolitan innovation, firm size, and business survival in a high-tech industry Alexandra ...

18 Pages·2014·0.58 MB·English

Metropolitan innovation, firm size, and
business survival in a high-tech industry
Alexandra Tsvetkova, Jean-Claude Thill
& Deborah Strumsky
Small Business Economics
An Entrepreneurship Journal

ISSN 0921-898X
Volume 43
Number 3

Small Bus Econ (2014) 43:661-676
DOI 10.1007/s11187-014-9550-z
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media New York. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
1 23
Author's personal copy
SmallBusEcon(2014)43:661–676
DOI10.1007/s11187-014-9550-z
Metropolitan innovation, firm size, and business survival
in a high-tech industry
Alexandra Tsvetkova • Jean-Claude Thill •
Deborah Strumsky
Accepted:30January2014/Publishedonline:22February2014
(cid:2)SpringerScience+BusinessMediaNewYork2014
Abstract This paper contributes to the growing entrepreneurstoshutdownexistingventuresinorder
body of business survival literature that focuses on topursueotheropportunities.
regional determinants of the hazard faced by firms.
Usingparametricsurvivalanalysis,wetesttheeffects Keywords Businesssurvival(cid:2)Metropolitan
of regional innovation on exit likelihood in the US innovation(cid:2)Survivalanalysis(cid:2)Computerand
computer and electronic product manufacturing dur- electronicproductmanufacturing
ing the 1992–2008 period. The novelty of our
approachisinconditioningtheeffectsofmetropolitan JELClassifications C41(cid:2)L26(cid:2)L63(cid:2)R1
innovation on firm size. Estimation results suggest a
negative relationship between metropolitan patenting
activity and survival of firms that started with 1–3
employees. This effect decreases if companies grow. 1 Introduction
Establishmentswithmorethan4employeesatstart-up
are insensitive to metropolitan innovation, although Empirical research shows that start-ups contribute
sizeoffirmsthatstartedwith4–9employeesimproves more than incumbent firms to job creation (Acs and
their survival chances. These findings indicate that Armington 2004). Also, their role in technological
localknowledgespilloversdonottranslateintolower evolutioniscrucialinthelongrun(FritschandMueller
hazard. The negative relationship indicates either a 2004).Manyfirms,however,exitsoonafterentry,as
creative destruction regime or decisions of only about half of them survive beyond 5 years
(Audretsch and Mahmood 1995; Dunne et al. 1989;
Johnson2005;Mataetal.1995).Exitofafirmdoesnot
A.Tsvetkova(&) necessarilyimplyafailure(Bates2005;Esteve-Perez
SchoolofPublicPolicy,GeorgeMasonUniversity, etal.2010).Mergersandacquisitions(M&A)areoften
3351FairfaxDr.,Arlington,VA22201,USA
regarded as a successful exit. Likewise, a company
e-mail:[email protected]
may decide to exit in order to pursue other business
J.-C.Thill(cid:2)D.Strumsky opportunities. Regardless of the business success an
DepartmentofGeographyandEarthSciences,University exitmightindicate,spendingpublicdollarstofacilitate
ofNorthCarolinaatCharlotte,9201UniversityCity
theestablishmentofshort-livedcompaniesislikelyto
Boulevard,Charlotte,NC28223,USA
e-mail:[email protected] be tantamount to an inefficient use of resources. For
thisreason,knowledgeoffactorscontributingtofirm
D.Strumsky
e-mail:[email protected] survivalisofpracticalimportanceforpolicy-makersas
123
Author's personal copy
662 A.Tsvetkovaetal.
they strive to implement programs designed to stim- survival likelihood in selected US industries as
ulate economic activity (Renski 2011). The desired proposed by agglomeration theory. He explains the
outcomesofsuchpoliciesareoftencouchedintermsof positiverelationshipbetweenthesethreedeterminants
increased firm formation and, most importantly, suc- andthehazardfacedbyfirmsinseveralindustriesby
cessfulperformanceofstart-ups. knowledgespillovers,amongotherfactors.
Perhaps for this reason, firm survival has become a Tocontinueontheimportanceofknowledge,recent
populartopicofresearchinthelastdecades.Theextant growththeoriespostulatethatinnovationandassociated
businesssurvivalliteraturepresentsadequateexplanation knowledgeandtechnologicalspilloversdetermineeco-
ofhowacompany’sinternalcharacteristicsandactivities, nomic growth in general, and firm performance in
andthenatureoftheindustryandrelevantmarkets,affect particular.Knowledgeisthemainstrategicresourcethat
itssurvivalchances.Instarkcontrast,ourknowledgeof a firm has at its disposal in the modern economy
the regional determinants of firm exits remains rather (Spender1996).Abusinesscaneithergenerateinnova-
limited (Brixy and Grotz 2007; Manjon-Antolın and tiveknowledge via R&D and other practices,orlearn
Arauzo-Carod2008).Althoughincreasingattentionhas fromotherfirms,orboth.Learningfromothersisaform
recentlybeenpaidtogeographicalfactors(Buehleretal. of knowledge spillovers, which increase productivity,
2012;Renski2011;Tsvetkovaetal.2014),apronounced mayleadtoincreasingreturnstoscale(Griliches1992;
gapinunderstandingtherelationshipbetweenaregional Lo´pez-Bazo et al. 2004), and therefore enhance the
milieu and business survival remains. As most policies chances of survival. The empirical literature suggests
are regional in nature, knowledge of the degree of that small firms are more likely to rely on knowledge
leveragethatpolicy-makersmayhaveonissuesoffirm spillovers,becausetheirownabilitytoinvestresources
survivalisessentialforeducatedpolicydesign.Ifregional inknowledgeproductionisusuallylimited(Audretsch
attributes are confirmed to systemically influence firm andVivarelli1996).Thisobservationimpliesthatone
survival,regionaldevelopmentpoliciescanbecraftedto should expect a positive effect of regional innovative
selectively stimulate the flourishing of attributes favor- environmentonbusinesssurvivalasknowledgeismore
abletosurvival,oralternativelytoencouragestart-upsto easilyacquiredfromnearbyinnovators.
locate in regions where auspicious conditions prevail. Conversely, a regional innovative environment is
This paper contributes to the business literature and intrinsicallyrelatedtotheso-called‘creativedestruction’
policy discourse by investigating the effects of one regime.1Firmsareusuallyinnovativeduringtheinitial
regionalcharacteristic,namelymetropolitaninnovation, stagesoftheirdevelopment,whentheytrytofindtheir
on survival of high-technology firms that did not nicheinthemarket.Noveltyhelpsnewfirmssucceedin
experienceM&A. competing against incumbents. By driving incumbents
We study computer and electronic product manu- to reinvent themselves or to go out of business,
facturing (NAICS334), as this industry critically entrepreneurialcompaniesachievesomemarketpower,
depends on innovations and should be particularly whichdoesnotstimulateinnovation.Afterafirmstops
perceptive to innovative environment. This industry creating new combinations and settles into running its
employs a disproportional number of researchers and businessjustlikeothers,itlosesentrepreneurialcharac-
engineerswhosegoalistodevelopnewproductsandto terandislikelytoendupexiting,whilenewinnovative
improve existing ones (BLS 2009).At the same time, firms keep introducing new combinations to the econ-
understanding business survival dynamics in NA- omy (Schumpeter 1942). The process by which less
ICS334 is important due to the industry’s special role entrepreneurialfirmsaredrivenoutofbusinessbymore
in the US economy. The computer and electronic entrepreneurial ones is the nature of ‘creative destruc-
product manufacturing industry promotes innovation, tion’.Itisourcontentioninthispaperthat,alongthisline
provides high-wage employment that has remarkable
multipliereffects,andaccountsformorethan10 %of
valueaddedinmanufacturing. 1 MarxandEngelscoinedtheterm‘creativedestruction’inThe
In the most complete account of the relationship CommunistManifestofirstpublishedin1848.Here,weusethe
Schumpeterianperspectiveoncreativedestruction(Schumpeter
between regional external economies and business
derived his notion of ‘creative destruction’ from the Marxist
survivaltodate,Renski(2011)carefullyexaminesthe
definition), which has a different meaning from the one
impactsoflocalization,urbanization,anddiversityon originallyproposed.
123
Author's personal copy
Metropolitaninnovation,firmsize,andbusinesssurvival 663
ofreasoning,moreinnovativeregionsshouldexperience have started to pay increasing attention to the
greater ‘creative destruction’, i.e. more firms entering geographic determinants of firm longevity (Brixy
and exiting the market, which in turn diminishes the andGrotz2007;Buehleretal.2012).Thejustification
likelihoodofsurvival. for linking regional characteristics to economic out-
Lastly, research-intensive regions may offer plenti- comesingeneral,andfirmperformanceinparticular,
ful business opportunities lacking in less innovative may come from at least four theoretical frameworks,
metropolitan areas. If business owners believe that namely the new economic geography (Fujita and
shuttingdownexistingbusinessesinordertore-channel Krugman 2004; Fujita et al. 1999; Fujita and Thisse
theresourcestostartanewventurewouldgiveagreater 2009), agglomeration theory that suggests local
return, one should expect a positive relationship knowledge spillovers (Jaffe and Trajtenberg 1996;
betweenregionalinnovationandexitlikelihood. Jaffeetal.1993;Marshall1920[1890]),clustertheory
In this paper, we empirically test these two (Porter 1990, 1998a, b), and the regional innovative
contrasting perspectives on the possible effects of systems perspective (Rodriguez-Pose and Crescenzi
regionalinnovationonbusinesssurvivalusingdatafor 2008; Uyarra 2010). In most general terms, the
the US computer and electronic product manufactur- relationshipbetweenregionalcharacteristicsandeco-
ing industry during the 1992–2008 period. We use nomic performance postulated by these frameworks
parametricsurvivalanalysistoconductthisstudy.The relies on two types of factors: those related to the
noveltyofourapproachisinconditioningtheeffects environment for doing business, such as specialized
of regional characteristics (metropolitan innovation) suppliersandappropriatelaborpool,andknowledge-
on firm attributes (start-up and current size). The related factors. The latter group appears to be more
empiricalliteratureshowsthattheeffectsoffirmsize prominentinthediscussionofregionaleffects.Thisis
may depend on the product and industry life cycle hardly surprising given that knowledge is nowadays
(Agarwal 1997; Agarwal and Audretsch 2001; Agar- the main strategic business resource (Spender 1996),
wal and Gort 2002; Agarwal et al. 2002), while the which is only partially excludable (Romer 1990) and
interaction of firm size with regional factors has not tends to spill over on businesses within some geo-
yetbeenstudied. graphic boundaries (Adams and Jaffe 1996; Bottazzi
Therestofthispaper isorganizedasfollows. The and Peri 2003; Rodriguez-Pose and Crescenzi 2008;
nextsectionbrieflyreviews theliteratureon business Wangetal.2004).
survival, agglomerated economies and knowledge Agglomeration theory to a varying degree incor-
spillovers, as well as other relevant perspectives. porateselementsfromtheotherthreeframeworksand
Section 3 presents the computer and electronic prod- isthemostrelevantforourresearch.Onemayexpect
uct manufacturing industry as an industry in which companiestofacealesserhazardingeographicareas
either of the two main perspectives on business withgreateraccumulatedstockofeconomicallyuseful
longevity2 may hold. Section 4 presents data sources knowledge(AcsandPlummer2005)andclosespatial
and the sample used in this study. The estimation proximity because local knowledge spillovers (LKS)
approachandvariablesaredescribedinSects.5and6, should increase productivity as a result of increased
followed by estimation results in Sect. 7. The last innovation and adoption of new technologies (Feld-
section contains concluding remarks and proposes man and Audretsch 1999; Jaffe et al. 1993; Koo
avenuesforfurtherresearch. 2005a). Besides, knowledge spillovers provide infor-
mationaboutbusinessopportunitiesinaregion(Porter
1998b), thus reducing the likelihood of an entry
2 Literaturereview mistake,apossiblereasonforexit(Jovanovic1982).If
learningandknowledgespilloversoccur,firmslocated
Traditionally, the empirical literature on business in the areas with intensive innovative activities, and
survival has focused on firm- and industry-specific formally not engaged in R&D, should become more
characteristics. In recent years, however, researchers efficient and innovative, face less uncertainty, and
enjoyhigherlikelihoodofsurvival.
2 We use the terms ‘‘business longevity’’ and ‘‘business A drastically different perspective suggests a neg-
survival’’interchangeablythroughoutthepaper. ativerelationshipbetweeninnovationinaregionand
123
Author's personal copy
664 A.Tsvetkovaetal.
onbusinesssurvival References Headd(2003),NafzigerandTerrell(1996),Persson(2004) Saridakisetal.(2008) ArribasandVila(2007),Headd(2003),Wilbon(2002) ColomboandGrilli(2007),Headd(2003),SantarelliandVivarelli(2007),Saridakisetal.(2008) NafzigerandTerrell(1996),Persson(2004) ArribasandVila(2007) Aspelundetal.(2005),Headd(2003),Littunen(2000) FontanaandNesta(2010),LinandHuang(2008) ´˜AgarwalandGort(2002),Esteve-PerezandManez-Castillejo(2008),Esteve-Perezetal.(2004),KaniovskiandPeneder(2008),Nikolaeva(2007),Fackleretal.(2013) Agarwaletal.(2002),Buddelmeyeretal.(2010),Levitasetal.(2006) Audretschetal.(2000),Fackleretal.(2013),FotopoulosandLouri(2000),KaniovskiandPeneder´(2008),Levitasetal.(2006),Mataetal.(1995),Persson(2004),SegarraandCallejon(2002),Strotmann(2007) ´˜Audretsch(1991),Esteve-PerezandManez-Castillejo(2008),Esteve-Perezetal.(2004),FontanaandNesta(2010),HuergoandJaumandreu(2004) BoyerandBlazy(2013),Buddelmeyeretal.(2010),ReidandSmith(2000) Audretschetal.(2000),FotopoulosandLouri(2000) ´˜Belloneetal.(2008),Esteve-PerezandManez-Castillejo(2008),Esteve-Perezetal.(2010),´FotopoulosandLouri(2000),SegarraandCallejon(2002) ´˜Esteve-PerezandManez-Castillejo(2008),Esteve-Perezetal.(2004,2010) BayusandAgarwal(2007),Buenstorf(2007),FontanaandNesta(2010),Persson(2004) JainandKini(2000) BridgesandGuariglia(2008),FotopoulosandLouri(2000),Headd(2003),MussoandSchiavo(2008),Saridakisetal.(2008) AudretschandMahmood(1995),LinandHuang(2008) ´LinandHuang(2008),SegarraandCallejon(2002) Belloneetal.(2008),Strotmann(2005) ´Belloneetal.(2008),KaniovskiandPeneder(2008),Mataetal.(1995),SegarraandCallejon(2002) ıAgarwalandGort(2002),Manjon-AntolnandArauzo-Carod(2008)
s
c
sti
eri
ct
chara cence)
vel oles
e d
theeffectsoffirm-andindustry-l Relationshiptofirmsurvival anagerialteam/keyemployees Positive Insignificant Positive Positive Negative Insignificant Positive Positive(liabilityofnewness) Positive,thennegative(liabilityofa Insignificant Positive(liabilityofsmallness) Positive Negative Positive Positive Positive Positive Positive Negative Negative Positive Negative Positive Negativeforyoungfirms
n m
o
h or
arc ner
e w y
Table1Summaryofres Characteristic Characteristicsofbusinesso Age Relevantexperience Education Diversebackground Firmcharacteristics Age Size Innovation Capitalintensity Productivityandprofitabilit Exporting Beingasubsidiary/branch Venturecapitalbacking Financialconstraints Industrycharacteristics Capitalintensity Marketpower Expandingindustries Matureindustries
123
Author's personal copy
Metropolitaninnovation,firmsize,andbusinesssurvival 665
business survival. This relationship may be deter- and in knowledge intensive industries in Europe, the
minedbyfactorsthatareexternalorinternaltoafirm. educationlevelofmanager(s)orowner(s)isfoundto
Competitive pressure is an external factor, while decrease hazard rates (Colombo and Grilli 2007;
decisionofabusinessownertodiscontinueoperation Headd 2003; Santarelliand Vivarelli 2007;Saridakis
forvariousreasonsisaninternalone. et al. 2008). On the other extreme, several studies
As argued by Schumpeter (1942), industries with report negative or insignificant relationship between
active entry are more innovative (entrepreneurial in educationalattainmentandsurvivalprospects(Arribas
his parlance). They should exert greater competitive and Vila 2007; Nafziger and Terrell 1996; Persson
pressure forcing incumbents, who are less likely to 2004). In fact, entrepreneurs with an education level
innovate, out of business. Indeed, empirical studies below average establish a substantial share of firms
showthatactiveindustryentrydecreasestheaverage (Christensen1997;DahlandReichstein2005).
business life expectancy (Kaniovski and Peneder Firmcharacteristicsthataffectbusinesssurvivalare
2008), while competition is more intense in highly inherently related to resource access. Predictably,
technological and innovative sectors (Agarwal and greater resourcefulness of a company enhances its
Gort 2002; Audretsch 1995; Segarra and Callejo´n survival chances. The empirical literature shows that
2002). This argument can be readily extended to size (Audretsch et al. 2000; Fackler et al. 2013;
regions in the case of industries that are dependent Fotopoulos and Louri 2000; Kaniovski and Peneder
predominantly on local markets either in component 2008)andage,atleasttoacertainpoint(Agarwaland
acquisition,orwhensellingthefinalproduct.Insuch Gort 2002; Fackler et al. 2013; Fontana and Nesta
instances one may speak of regions that differ in the 2010), reduce hazard. Other factors related to lower
intensity of ‘creative destruction’. Firms in more hazardarethefirmstatusofanexporter,asubsidiary,
innovative regions, particularly those not engaged in or a branch, and access to venture capital financing
research and development, are likely to be disadvan- (Bayus and Agarwal 2007; Esteve-Pe´rez and Man˜ez-
taged and face a higher hazard due to increased Castillejo 2008; Esteve-Perez et al. 2010; Jain and
competitivepressure. Kini 2000). Successful innovation, capital intensity,
Vibrantregionscharacterizedbyintensiveinnova- profitability,andproductivityusuallyleadtoalonger
tion may offer greater business opportunities. In this lifespan for a firm (Audretsch 1991; Bellone et al.
case, shutting down a company to pursue a more 2008; Fotopoulos and Louri 2000). In contrast,
promisingventuremaybetheoptimalstrategyforan financial constraints (Bridges and Guariglia 2008;
entrepreneur. For example, Bates (2005) finds that Headd 2003; Musso and Schiavo 2008), as well as
alternativebusinessopportunitiesareakeyreasonfor costs of innovation that are beyond a company’s
businessclosureamongsmallyoungcompaniesinthe means(BoyerandBlazy2013),increasethelikelihood
US ofexit.
Inadditiontoresearcheffortsdevotedtothestudy Industrial characteristics shown to influence busi-
of geographical determinants of business survival, nesssurvivalincludecapitalintensity,marketpower,
extensive literature investigates firm- and industry- and the stage of industrial life cycle. A positive
level factors that affect firm longevity. Table 1 relationship between capital intensity and firm lon-
succinctly presents current knowledge in this area gevity reported by several studies (Audretsch and
followedbyabriefdiscussion. Mahmood 1995; Lin and Huang 2008) may be the
Human capital available to a firm is perhaps the result of self-selection when more resourceful firms
most decisive performance factor among individual enter capital-intensive industries in a hope of less
characteristics. Relevant experience ofan owner ora competition (Doms et al. 1995). Lower competition
managerconsistentlydecreaseshazardfacedbyafirm promotes(LinandHuang2008;SegarraandCallejo´n
(Arribas and Vila 2007; Headd 2003; Wilbon 2002). 2002) or hampers (Bellone et al. 2008; Strotmann
The size and diverse backgrounds of the co-founders 2005) business survival, depending on specific cir-
or a management teams usually translate into cumstances. Expanding industries usually provide
increased probability to stay in business (Aspelund favorable conditions for both incumbents and new
etal.2005;Headd2003;Littunen2000).Theevidence entrants, which tend to stay in business longer
ontheeffectsofeducationiscontroversial.IntheUS, compared to the companies in stable or contracting
123
Author's personal copy
666 A.Tsvetkovaetal.
sectors (Bellone et al. 2008; Kaniovski and Peneder emphasis on R&D in their day-to-day operations
2008;Mataetal.1995;SegarraandCallejo´n2002). (BLS2011).
In addition, despite the fact that the computer and
electronic product manufacturing industry is global
3 Computerandelectronicproduct when it comes to selling finished products, the
manufacturingindustry intermediate production stages tend to be local.
Computers and electronics contain multiple compo-
Industriesdiffersubstantiallyintermsofpropensityto nents,producedbydifferentcompanies.Accordingto
exit(Dunneetal.1989),possiblyduetothedifferences the Bureau of Labor Statistics, NAICS334 firms that
in local knowledge spillovers intensity (Glaeser et al. produce intermediate components and finished pro-
1992),competitionregime(Fritschetal.2006;Segarra ducts prefer to locate in close proximity in order to
and Callejo´n 2002), minimum efficient scale (Fritsch enjoy immediate access to innovations (BLS 2009).
etal.2006),firms’absorptivecapacity(Fabrizio2009), The local dimension of the industry makes both the
and other characteristics. In this paper, we focus our knowledge spillovers argument and the ‘creative
analysisoncomputerandelectronicproductmanufac- destruction’ hypothesis pertinent as two possible
turing.3Althoughtheresultsofthisresearchmaynotbe explanations for the relationship between metropoli-
readilygeneralizabletoothermanufacturingindustries taninnovativeenvironmentandfirmsurvival.
because of such restriction, understanding business
survivaldeterminantswithintheNAICS334industryis
important due to the significance of computer and 4 Dataandsample
electronicsmanufacturingfortheUSeconomy.Recent
research shows that this industry provides high-wage The National Establishment Time Series (NETS)
employment and plays an important role in industrial Databaseforyears1991–2008isthemaindatasource
innovation, trade deficit reduction,environmentalsus- usedforthisanalysis.TheNETSDatabaseiscreated
tainability (Helper et al. 2012), and growth (Koo by Walls and Associates from the Dun and Brad-
2005b). On average, NAICS334 accounts for about street’s(D&B)DUNSMarketingInformationarchive.
11 % ofall the value added bytheUSmanufacturing The database consists of yearly snapshots of the US
(approximately 1.7 % of GDP)4, and its multiplier economy (all firms recoded by D&B to be active)
effectacrosstheeconomyissubstantial.5 performedeveryJanuary since 1990.Thedatabase is
Computerandelectronicproductmanufacturingis updatedeverysummer.Ifanestablishmentgoesoutof
well suited for the study of external effects of business,itslastyearofoperationisindicatedbutthe
innovation on business survival for several reasons. record is not removed. This allows for the study of
First,itisanindustrythatconsistentlyranksasoneof activecompaniesandestablishmentsthathaveexited.
themostinnovative.Itssuccessislargelybasedonthe TheNETSfileavailableforthisresearchisasubsetof
development and introduction of new products, tech- the original database. The file contains longitudinal
nologies, and software. There is a high pressure on informationabouteachestablishmentstartedin1991,
NAICS334 companies to innovate, thus explicit includingcompanyname,countyFIPScode,yearsof
operation (first and last years in the dataset, year the
business started), industry classification (6-digit NA-
ICScode),typeofestablishment(standalone,branch,
3 TheindustryconsistsofNAICS3341(computerandperiph-
headquarter), and the number of employees. We
eralequipmentmanufacturing),NAICS3342(communications
supplement the NETS Database with data available
equipment manufacturing), NAICS 3343 (audio and video
equipment manufacturing), NAICS 3344 (semiconductor and fromtheUSPatentandTrademarkOffice(PTO),the
other electronic component manufacturing), NAICS 3345 CensusBureau,theBureauofLaborStatistics,andthe
(navigational, measuring, electromedical, and control instru-
IntegratedPostsecondaryEducationDataSystem.
ments manufacturing), and NAICS 3346 (manufacturing and
Foroursample,weidentifyallestablishmentsthat
reproducingmagneticandopticalmedia).
4 http://www.bea.gov/industry/gdpbyind_data.htm. werestartedinyear1991andtrackthemtill2008,the
5 Forinstance,themultipliereffectofcomputermanufacturing last year of data availability. To mitigate a potential
isestimatedtobe16inCalifornia(DeVoletal.2009). risk of aggregation bias we use only stand-alone
123
Author's personal copy
Metropolitaninnovation,firmsize,andbusinesssurvival 667
Table2 Total number of NAICS334 start-ups in 1991, and facedbyacompanyoveryearsisnotindependent;all
establishmentsintheestimationfile observations for the same firm share frailty, which is
Descriptiona Numberoffirms assumedtofollowagammadistribution.6
We first perform the analysis using all establish-
Totalnumberofstart-upsin1991 2,658 ments in our sample to depict the overall pattern of
OutsideofcontinentalUSA 11 effects on business survival. In the next step, we
OutsideMSAs 229 estimate the same model separately on three sub-
Notindependent 261 samples of firms determined by the number of
ExperiencedM&A 12 employeesatstart-up.Theadditionalanalysesprovide
Havemorethan100employeesin1992 6 adeeperunderstandingoftheconditioningroleplayed
Outliers 1 byfirmsizeontherelationshipbetweenmetropolitan
Missing/erroneousdatainNETS 385 innovationandlongevity.
Totalestablishmentsinthesample 1,803 The empirical literature often distinguishes com-
panies with one to three employees (Fackler et al.
SourceNETSDatabase,USPTO,TheDealPipeline,WRDS,
AlacraStore 2013), one to nine employees (Arvanitis and Stucki
a Thesecategoriesarenotexclusive 2013; Uhlaner et al. 2013), or up to 25 employees
(RosenthalandStrange2003)asseparategroups.We
establishments in US MSAs in our analysis; retained divide our sample into three groups. The first group
establishments also had less than 100 employees in containsbusinessesthathadonetothreeemployeesin
1992, and did not experience merger or acquisition. 1992 (798 firms); the second group includes compa-
Table 2 presents all categories of firms that were nieswithfourtonineemployeesin1992(630firms),
removed from the estimation file and firm count by and all other firms constitute the third group (375
category. The final sample consists of 1,803 estab- firms). For the purposes of this analysis, we refer to
lishments,1,122ofwhichexitedovertheobservation these groups as small firms, medium firms, and large
periodasshowninFig. 1. firms,respectively.
5 Econometricanalysis 6 Explanatoryandcontrolvariables
We use parametric survival analysis to estimate the Forsurvivalanalysis,onlyindependentvariablesneed
effects of metropolitan innovation and firm size on tobespecified,asthedependentone,thehazardrate,is
business survival in the US computer and electronic estimated implicitly. The main explanatory variables
product manufacturing industry. The validity of para- in this study are the level of metropolitan innovation
metric estimation results critically depends on the (Innovation),firmsize(Size),andaninteractionterm
appropriate exit distribution that is hypothesized. The between Innovation and Size (InnovSize). Innovation
literaturesuggestsahighlikelihoodoffirmexitduring in a region may be measured in a number of ways.
the first years in business, which decreases over time Patenting intensity, R&D expenditures, or share of
(Fackler et al. 2013). The smoothed hazard estimates R&Demployeesarecommonlyusedintheliterature.
(Fig. 2)andthedynamicsofexitsinoursample(Fig. 1) Yet, it is not the patents, or R&D expenditures, or
exhibitthesamepattern.Forthisreason,weusethelog- shareofR&Demployeesintheeconomythatdirectly
logisticdistribution,whichallowsforinitiallyincreas- increase productivity and make regional economies
ingandsubsequentlydecreasinghazard. more innovative and prosperous. The main driver of
Businesssurvivalisafunctionofmultiplefactorsat economic success demonstrated by companies and
variouslevels.Individualcompanycharacteristicsare regional economies is the fruitful application of new
likely to play a central role in the ability of a firm to ideas in the market. However, new ideas cannot be
live longer. Often such characteristics are not measuredasthey do notleave a paper trail,torepeat
observabletoaresearcher.Toaccountforunobserved
heterogeneity at the establishment level, we fit a 6 Modeling frailty as gamma distributed is standard practice
shared frailty model. Under this specification, hazard (Hougaard1995).
123
Author's personal copy
668 A.Tsvetkovaetal.
Fig.1 Numberof
survivingestablishmentsin
thesampleoveryearsand
exitdynamics
unpatented. On the other hand, not every patent is
utilized in the market promoting productivity and
innovation. Furthermore, the number of patents as a
measureofinnovationisunabletoreflectmarketvalue
of each patent, which are likely to differ greatly
(Levitas et al. 2006). Despite all these concerns,
researchers have argued that patent count is perhaps
the best measure of innovation (Griliches 1990),
because it is superior to other available measures
(Feser 2002)and is anappropriate approximation for
the stock of knowledge generated in an urban region
(Acsetal.2002).
Firmsizeisarguablythemoststudieddeterminant
ofbusiness survival.Numerous studies report the so-
Fig.2 Smoothed hazard estimates for the US computer and
called ‘‘liability of smallness’’, a consistent positive
electronicproductmanufacturing,1992–2008
relationship between the size of a company and its
longevity(Audretschetal. 2000;Fackleretal.2013;
Krugman’s (1991) famous saying. Thus, researchers
Fotopoulos and Louri 2000; Kaniovski and Peneder
havetouseapproximationsforthestockofknowledge
2008; Levitas et al. 2006; Mata et al. 1995; Persson
generatedinaregion.
2004; Segarra and Callejo´n 2002; Strotmann 2007).8
We use the natural logarithm of the population-
Researchers use various metrics to approximate the
adjustedtotalnumberofsuccessfulpatentapplications
size of a business, such as the number of employees
in a given year in a given MSA as a measure of
(currentoratstart-up),assets,orsalesvolume.There
innovation.7 The US PTO is the data source. As a
is some evidence that the results are in general not
proxy for the stock of new profitable ideas, patent
sensitivetotheapproximationchosen(Agarwaletal.
count has its weaknesses, as well as limitations.
2002), although the current number of employees is
Patents do not cover all new ideas created in the
preferred as a measure of firm size in the business
economy; possibly the majority of them actually go
survivalliterature(Mataetal.1995).Inthisstudy,Size
is the natural logarithm of the current number of
7 Evaluation of a patent application is a lengthy process; in employees as reported in the NETS Database. Vari-
addition,theremaybeadelayintheUSPTOsystemreporting ableInnovSizeiscalculatedbymultiplyingInnovation
ofthenumberofpatentsgranted.Arguablyforthesereasons, bySize.
patentcountsreportedbytheUSPTOatthetimethedatawere
retrieved for this study decrease sharply for the last 3years
covered by the analysis. To mitigate a potential bias in
estimation due to the measurement error we inflate patent 8 Several studies, however, report statistically insignificant
countsintheyears2006,2007,and2008by5,10,and15%, effectofsizeontheprobabilityofexit(Audretschetal.2000;
respectively. Saridakisetal.2008).
123