Forschungsinstitut
zurZukunft der ArbeitInstitute for the StudyofLabor
ThePostCrisisGrowthintheSelf-Employed:VolunteersorReluctantRecruits?
Andrew Henley
AberystwythUniversityandIZA
DiscussionPaperNo.9232July2015
IZA
P.O. Box 724053072 Bonn
Germany
Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail:
Anyopinionsexpressedherearethoseoftheauthor(s)andnotthoseofIZA.Researchpublishedinthisseriesmayincludeviewsonpolicy,buttheinstituteitselftakesnoinstitutionalpolicypositions.The IZA research networkis committed to the IZA Guiding Principles of Research Integrity.
TheInstitutefortheStudyofLabor(IZA)inBonnisalocalandvirtualinternationalresearchcenterandaplaceofcommunicationbetweenscience,politicsandbusiness.IZAisanindependentnonprofitorganizationsupportedbyDeutschePostFoundation.ThecenterisassociatedwiththeUniversityofBonnandoffersastimulatingresearchenvironmentthroughitsinternationalnetwork,workshopsandconferences,dataservice,projectsupport,researchvisitsanddoctoralprogram.IZAengagesin(i)originalandinternationallycompetitiveresearchinallfieldsoflaboreconomics,(ii)developmentofpolicy concepts, and(iii) dissemination of researchresults andconcepts tothe interested public.
IZADiscussionPapersoftenrepresentpreliminaryworkandarecirculatedtoencouragediscussion.Citationofsuchapapershouldaccountforitsprovisionalcharacter.Arevisedversionmaybeavailable directly fromtheauthor.
IZADiscussionPaperNo.9232July2015
ABSTRACT
The Post Crisis Growth in the Self-Employed:Volunteers or Reluctant Recruits?
IntheUKbylate2014therewerealmost0.75mmoreself-employedthanatthestartofthefinancialcrisisinearly2008.Thisrepresentsover75%ofjobsgrowthintheUKoverthesameperiod.Thisexperiencehasattractedcommentaryfromindependentpolicyanalystsandothers,focusingonwhethergrowthhasbeenstructural,reflectingchangesinthenatureofemploymentandattitudestowardsbusinessventuring,orcyclical,reflectingapost-crisisshifttowardsflexibleinsecureformsofemploymentasanalternativetolong-termunemployment.RecentcommentaryhasalsofocusedonheterogeneityacrossUKregions.Longitudinaldatacovering2009-2013fromtheESRCUnderstandingSocietysurveyareusedtoexaminetransitionsintoself-employment,andregressioncorrelationwithindicatorsoflabourmarketconditions(unemployment,earnings)inthearealocaltotheindividual.Transitionsintoself-employmentfrombothpreviouspaidemploymentandinactivityfoundtobearenegativelycorrelatedwithlaggedlocalunemploymentratesandpositivelycorrelatedwithlaggedlowerquartileearningsinthelocalarea.Thesecorrelationpatterns,althoughvaryinginsize,holdformenandwomen,andarerobusttocontrollingforindividualcharacteristics.Thissuggeststhatlocalpullfactorsarefarmoresignificantindrivingtransitionsinto self-employment, andexplains why businessformation ratesare higher, post-2008,inmoreadvantagedUKareas.Self-employedbusinessownershipdoesnotappeartoasignificantalternativetounemploymentforthosewhereofpaidemploymentdemandisweak.Entrepreneurialactivityprosperswerewagesarehigherandunemploymentlower.
JELClassification:J21,M13,R23
Keywords:self-employment,localunemployment,localearnings,longitudinalanalysis
Correspondingauthor:AndrewHenley
SchoolofManagementandBusinessAberystwythUniversity
LlanbadarnAberystwyth,SY233ALUnitedKingdom
E-mail:
1.Introduction
IntheUKby theendof2014therewerealmostthreequartersofamillionmoreself-employedintheworkforcethanatthestartoftheglobalfinancialcrisisinearly2008.ThisisaremarkablenumericalgrowthandrepresentsinturnoverthreequartersofthetotalnetgrowthinjobsintheUKoverthesameperiod(Table1).TheUKexperience,althoughdramatic,isnotuniqueacrossEurope.Highratesofgrowthofself-employmenthavebeenobservedinsomeoftheeasternEuropeanaccessionstatesandtoalesserextentinFranceandtheNetherlands.Nevertheless,acrosstheEUasawholeself-employmenthasfallenslightlyoverthisperiod.TheUKexperiencehasrecentlyattractedcommentaryfromindependentpolicyanalystsaswellasconcernedorganizationssuchastradesunions(D’ArcyandGardiner, 2014; Hatfield, 2015; Centrefor Cities,2015). Discussion hasfocusedontheextenttowhichthegrowthhasbeenstructural,reflectingsocialchangesinthenatureof employmentandattitudestowardsbusinessventuring,cyclical,reflectingapost-crisisshifttowardsflexiblepart-timeformsofemploymentasanalternativetolong-termunemployment,orspatial,reflectinggeographicalvariationinsupply-anddemand-sideinfluences.The present paperis concernedwith the last of these.
Althoughself-employmenthasrisenacrossallregionsanddevolvedterritoriesoftheUK,withtheexceptionofNorthernIreland,thescaleofthatgrowthhasvaried.Insomeregions,notablyinLondonandthesoutheastofEnglandself-employmentgrowthhasbeenmatched bygrowthinemployeejobs;inothers,notablyinthenorthoftheUK,ithasnot.Allofthisrecentexperiencemay situneasily withthechangeinresearchfocusoverrecentyearstowards viewing entrepreneurial choiceasdrivenbyinnovation and knowledge spillovers.Toparaphraseaninfamousquestionfromanearlierdebateaboutthedepressioncrisisofthe
1930s, is thearmyoftheself-employedstandingwatch overthecurrentfragilerecoveryfromfinancialcrisisavolunteeroneornot?1Arethosechoosingself-employmentreluctantlydoingsoinplaceswherethealternativesarenotveryattractive,ormakingactivechoicesinplaceswhere businessstart-upopportunities look more attractive?
Thedebatebetweenviewingself-employmentasopportunity-drivenornecessity-driven(intheterminologyoftheGlobalEntrepreneurshipMonitor)isofcoursenotanewone.Ithasattractedsignificantresearchattentioninthepast(forexampleGiladandLevine,1986;Amit,1994;Hesselsetal.,2008;Thuriketal.,2008;Dawsonetal.,2014).Byandlargepastresearchhaspointedtowards thedominanceofopportunityas amacro-leveldrivingforce.Thevastmajorityoftheself-employedappeartoreportopportunity-relatedorpersonalindependence-relatedmotivesfortheirchoiceofeconomicstatus,andnottoattributeanysignificanceto“recession-push”factors.Thispointstowardssomecoincidencebetweenself-employmenttrendsandentrepreneurialactivity,broadlydefined.Howeverdetailedmicro-econometricanalysis,whichwillbediscussedfurtherbelow,alsopointsto thesuggestionthatthosewhoformnewself-employedbusinessventuresfrompriorunemploymentmaynot achieve thesame levels ofgrowth or business longevityas others.
Itisthereforeimportanttoknowmoreaboutthe(newly)self-employedinpost-crisis“austerity”UnitedKingdom.Towhatextentisvariabilityinlocallabourmarketconditionsandthestateoflocaldemandassociatedwithindividualtransitionintoself-employment(stocksandinflows)?Andtherefore,perhapsmorespeculatively,whataretheimplicationsofthisfor thewayin whichpolicyto supportentrepreneurshipframed?
1The debate in question revolvedaround therelationshipbetweenunemploymentand thelevel of unemploymentinsurance(BenjaminandKochin,1979).
Inordertounderstandthedynamicsofself-employmentattheleveloftheindividual,itisessentialtoanalyzelongitudinaldata,ratherthansuccessivesurveysdrawnfromdifferentsamplesofthesamepopulation.Thispaperinvestigatesindividual-levelmicrodatafromthefirstfourwaves(2009-2013)ofUnderstandingSociety,theUK’sprincipalhouseholdlongitudinalsurvey,andundertakesdatalinkagetoOfficeforNationalStatisticsinformationonearningsandunemploymentfor380GreatBritainlocalauthoritydistricts.Itanalysestheextenttowhichself-employmentstatusandtransitions(frompaidemploymentandfrominactivity)intoself-employmentareassociatedwithvariationinpriorlocaleconomic conditionscapturedbyunemploymentratesandlevels ofearnings.
Thepaperconcludesthatthereislittleornoevidenceforanynet“push”effectintoself-employmentfromweaklocallabourmarketconditions.Thedataareconsistentwiththeneteffectbeingoneof“pull”inwhichimprovedlocallabourmarketconditionsindicatebetterlocalbusinessopportunitiesandmarketdemandfromhigherspendingpower.“Pull”effectsappeartobestrongerforwomenandstrongerstillforthoseconsideringatransitioninto self-employmentfrom inactivity.
Theremainderofthepaperisstructuredasfollows.Section2providesafullerbackgrounddiscussion,andreviewofbothrelevantacademicliteratureandrecentpolicy-focusedcommentary.Section3describesthedatausedandthemainmethodsofanalysis.Section4explainstheresults.Section5providesadiscussionandcommentaryonthesefindingsandtheirimplications.Section6isafinalconclusionand commentonpolicyimplications.
2.Regionaldrivers ofself-employment
Self-employmentcurrentlyaccountsforover15%ofallthoseinworkintheUK,oroverfourandhalfmillionindividualsfromaworkforceofalmost31million.Tosetthisincontext,steadygrowthinself-employmenthasbeenapronouncedfeatureofthedevelopmentoftheUKlabourmarketforsomeconsiderabletime.AsFigure1shows,35yearspreviouslyin1979therateofself-employmentwasexactlyhalftherateinmid2014.Growthoverthatperiodhasnotbeenuniform:theperiodofhighunemploymentinthe1980switnessedmorerapidgrowth,withsomedeclineintherateduringtheperiodofsustainedeconomicgrowththroughthe1990s.Theself-employmentratestartedtogrowmorerapidlyagainfromaround2008onwards,coincidingwiththeonsetoftheglobalfinancialcrisis.Figure2chartsmovementsinUKunemploymenttotalssince2000quarterbyquarter,alongsidethetotalnumbersofself-employed.Itisdifficulttodrawanyfirmconclusionfromthesuperimpositionofthesetwotimeseries–itdoesnotsuggestionanystrongpro-orcounter-cyclicaltrend.
Thepost-crisisdevelopmentoftheUKeconomyhasbeennotableinthattotalnumbersinworkhasmaintainedsustainedgrowth,withnodramaticgrowthinunemploymentthatcharacterizedearlierrecessionsintheUK,notablyinthe1930sand1980s.Table1,drawnfromofficialUKONSstatistics,describesthecreationofjobsintheUKbetween2008and2014andshowsclearlyhowthisnetgrowthinjobsissignificantlyaccountedforbythegrowthinself-employment.Of920,000netnewjobscreatedbetweenquarter1of2008andquarter2of2014,693,000wereinself-employment.Thenetfigurerepresentsabalanceofinflowsandoutflows.Inflowsaccountforover36%ofthetotalinself-employment five years previously; outflows represent only 23% of those in self-
employmentfiveyearspreviously.Thisamountstoasignificantgrowthineithersmall-scalebusinessventuringorfreelancing/own-accountself-employmentorboth(andatatimewhenUKtaxauthoritieshaveprogressivelytightenedrulesofself-employmentregistration,HMRC(2013)).Table1alsoreports,forcomparisonpurposes,datashowingthechangeinthepopulationofmicrobusinesses(0-10employees)intheUKoverthesameperiod.DataincludebusinesseswhichareregisteredforVATand,ifemployers,formakingtaxandnationalinsurancepayments(pay-as-you-earn,PAYE)onbehalfofemployees,aswellasunregisteredsole-traders.Althoughtheincreaseofover290,000netnewbusinessesisnotonthesamescaleasforself-employment,itisstillsignificant.D’ArcyandGardiner(2014)providecomparativeinformationonrecenttrendsinthesetwomeasures.Thedifferenceinthedatasuggestsanumberofpossibilities,includingsignificantgrowthinnon-businessownershipself-employment,orthetransitioningintofull-timeself-employmentbyindividualswhohadalreadyregisteredbusinessesorwerejoiningbusinessesalreadyestablishedbyother self-employmentindividuals.
Growthinself-employmenthasbeenproportionately higheramongstwomenovertheperiod2009to2014.InfactUKLabourForceSurveyestimatessuggestthatgrowthinfemaleself-employmentaccountedforalmosthalfofthetotalgrowth,eventhoughin2014men still accountfor 68%of allself-employed.
Theregionalpatterninthegrowthofself-employmenthasnotbeenuniformacrosstheUK.Table2describesthepatternsatthehighlevelofgovernmentofficeregion(NUTS1).Self-employmenthasgrownsignificantlyinallregionsanddevolvedterritories
withtheexceptionofNorthernIreland.2Onemighthypothesizethateitherself-employmentgrowthwouldbehighestinthefastestgrowingregionsclosesttoandincludingLondon,orthatself-employmentgrowthmighthavebeenstrongerintheweakerregionsmostatriskfromjoblossesfollowingthefinancialcrisis.Inthedatathreegroupsofnotnecessarilycontiguousregionsemerge.InLondon,theSouthEast,theWestMidlandsandtheNorthWestself-employmenthasgrownbetween17and19%overtheperiod2008to2013.IntheNorthEast,theEast,theSouthWestandinWalesthegrowthratehasbetween10and12%.InScotland,YorkshireandHumbersideandtheEastMidlandsgrowthhasbeenaround7to8%.Itisdifficulttomakeparticularsenseofthesegroups–excepttoproposethatself-
employmentgrowthinparticularregionsappearstoreflectacomplexbalanceofeconomicdemandandlabourmarketpressures,and,totheextentthatdevolvedterritoriesarefreetoadoptdifferentpolicyinstruments,variationinpolicyactivism.Nevertheless,intermsofabsolutenumbers,thegrowthinself-employmentisdominatedbygrowthinLondon,theSouth Eastand the North West.
Exantethetheoreticalrelationshipbetweenentrepreneurialactivityandunemploymentisambiguous.Economicanalysisofentrepreneurialchoicepredictsthathigherunemploymentwillinducemoretoswitchintoself-employmentbecauseincreasingunemploymentraisesthedifferentialbetweentheexpectedreturnfromusingone’shumancapitalinbusinessownershipandthatexpectedfromattemptingtodeploythathumancapitalinthepaidlabourmarket(Parker,2009).However,theregionaleconomicsliteraturearguesthatentrepreneurialopportunitiesarelessprevalentinlocalitieswhereunemploymentis
2TheNorthernIrelandexperiencehasbeensomewhatdifferenttothatoftherestoftheUK,largelybecauselevelsofself-employmenthavehistoricallybeenmuchhigherintheprovinceduetogreatersignificanceofagricultureandrelatedruralsectorstoitsregionaleconomy.Sincethe1998GoodFridayAgreementtheregionhasexperiencedasignificantmeasureofeconomicstabilitythathassupportedinwardinvestment,industrialrestructuringandgrowthof new sectors,andemploymentgrowth.
higher,becausehigherunemploymentcorrelateswithlowereconomicdemandandthereforebusinessopportunity(Storey,1991;Reynoldsetal.,1994).Thequestionoftherelativeimportanceof“prosperity-pull”and“recession-push”effectsisonethatdominatesextantresearchontherelationshipbetweenentrepreneurialactivityandunemployment,andfiguressignificantly in discussion on the spatial variation in self-employment and new firm
formation(seeAudretschetal,2014).3Ithasbeennotedonmanyoccasionsthatthereisno
clear-cutempiricalrelationshiphere,reflectingthetheoreticalambiguity(ArmingtonandAcs, 2002; Thuriketal.,2008; Parker, 2009; Audretschetal.,2014).
Self-employmentgrowthmayreflect“necessity”motives,asunemployedworkersintimesandinlocationsofweakerlabourdemand,turntoentrepreneurshipasanalternativesourceofearnings.Thus,governmentsmaypursueactivelabourmarketpoliciesthatencourageandsupportbusinessventuringasameansofalleviatingunemployment(Frisch,1993;BaumgartnerandCaliendo,2008;CaliendoandKunn,2014).Newbusinessownersmayfinditeasiertohireotheremployeeswhenunemploymentishigher,reinforcingthehigherunemployment-higherentrepreneurialactivityargument(Henley,2005).Ontheotherhandself-employmentgrowthmayreflectwiderimprovementsineconomicdemandforgoodsandservicesatparticulartimesandlocations.Thepaymentofredundancycompensationmayalsoencouragetransitionintoself-employmentbytheunemployedbecausewindfallpaymentsmayprovideaready sourceofbusinesscapital(LindhandOhllson, 1996; Taylor, 2001; HurstandLusardi,2004; Georgellisetal., 2005).
Parker(2009)notesthatearlycrosssectionevidencetendstosupporttheconclusionof a negative association (“prosperity-pull”) between local unemployment rates and the
3 Audretsch et al. (2014), Table 1, provides a helpful summary of previous empiricalresearch.
probabilitythatanindividualwouldbeinself-employment.Howevermorerecentresearchappearstoquestionthatconclusion(seeParker,2009,Table4.1,p.108forameta-analysis).InsteaditsquareswiththeobservationofEvansandLeighton(1990)thatahigherproportionoftheunemployed,whencomparedtothealreadyemployed,appeartotransitionintoself-employment.Researchonthestrengthoftheassociationbetweenunemploymentandnewfirmformationreinforcesthisambiguity,butdoesfindclearevidenceofapositiveassociationbetweenbusinessventuringandlocaleconomicprosperity(FrischandStorey,2014).
Atleastonerecentmicro-econometricanalysis,inthiscaseforolderAmericans,supportsthefindingthattheunemployedaremorelikelytoenterself-employment(Biehletal.,2014).Thisrecentstudyservestoreinforceanimportantpoint–namelythatmuchpreviousresearchhasfocusedontherelationshipbetweenunemploymentandthesizeoftheself-employed“stock”.Giventhatchoices,particularbusinessinvestmentdecisions,maybesubjecttoinertia,itismoreimportanttofocusontransitionsintoself-employment.Observedstatusmaybeassociatedwithdrivingfactorsatsomevariablepointinthepastdependingonthelengthofanycurrentself-employmentspell.Cross-sectionalcorrelationsbetweenlevelsofself-employmentandlocalunemploymentratesmaynotrevealmuchatallaboutpastdrivers.Transitiondecisionsaremorelikelytobeinfluencedbyobservedconditionsimmediatelyprecedingthe decision.
Aninvestigationoftransitions(flows)ratherthanself-employmentstatus(stocks)normallyrequireslongitudinaldata.Furthermorelongitudinaldatamayallowentrepreneurialchoicetobemoreaccuratelydefinedandmodelled(FrischandStorey,2014).However,evenwithin longitudinal data sources such as household panel surveys the sample numbers
transitioningintoself-employmentmaybequitesmall,particularlyforthosetransitioningfromunemployment,makingitdifficulttoidentifystrongassociationswithparticulardriverfactors.Longitudinaldataalsoallowstheimplicationsofenteringself-employmentfrompriorunemploymenttobe investigated.Survivalrates(spelllengths)inself-employmentmaybelower(Millánetal.,2012).Businessesstartedbytheunemployedmayalsonotgrowasquickly(HinzandJungbauer-Gans, 1999; Caliendo andKunn,2014).
Timeseriesanalysesmeasuringthestrengthofanyassociationbetweenself-employmentandunemploymentaggregates,eitherateconomy-levelorlargeregional-level,providecontrastinmethodologicaltermswiththesecross-sectionalandlongitudinalindividual-levelanalysis.AnauthoritativestudyisThuriketal.(2008)whichconcludeswithevidenceforbothpushandpulleffects,butthatthe“prosperity-pull”hypothesisismuchmoresignificantinquantitativeterms.Otherregionalpanelworkalsoprovidessupportforsome degree of “recession-push” impact onself-employmentrates (Robson, 1998; Parker andRobson,2004;NittykangasandTervo,2005;Tervo,2006).There maybeasimilarimpactonregionalnewfirmformationrates(Ritsilä andTervo,2002).
Theoreticalambiguityexistsbothbetweenandwithinthenotionsof“prosperity-pull”and“recession-push”.Inessencebehindthe“recession-push”hypothesisaretwofurthertwoideas.Thefirstisthathighernumbersofunemployedinalocalitymayresultinagreaterlikelihoodthatthoseenteringself-employment dosofromunemployment,andthatthislikelihoodincreasesfurtherbecausepaid-employmentjobsearchismoredifficultandcostly.Thesecondisthathigherlocalunemploymentisassociatedwithgenerallymoredifficultlocallabourmarketconditions.Inthecontextoftheaftermathofthe2008crisisintheUK,thatmeantformostworkersverylowratesofanticipatednominalwagegrowth(andreal
wagedeclines),andforsome,whereemployerswereabletonegotiatepaycutsinordertopreservejobs,verysignificantnominalandrealdeclinesinearnings(Greggetal.,2014).Furthermore,perceivedjobsecuritymayhavefallensharply,and,underpressurefromemployerstorestorefallingproductivitylevels,jobsatisfactionmayhavefallenaswell(McManus,2012,commentingontrendsintheBritishSocialAttitudesSurvey).Allofthesepressuresmayservetoraisetheperceivedrelativeattractivenessofself-employment.Ofcourse,inreality,actualself-employmentmayalsoturnouttobeunrewarding,unsatisfyingandinsecure (D’Arcyand Gardiner, 2014; Hatfield, 2015).
Audretschetal.(2014)makeanimportantobservationthatcurrentlocalunemploymentratesmaskheterogeneityaboutlocallabourmarketconditions,inparticulartheextenttowhichlocalvariationintheskilllevelsoftheunemployed,aswellasvariationinspellduration,revealsinformationabouttheirabilitytocompeteforavailablejobs,andthereforeexperiencelower“push”towardsself-employment.Recentpost-crisisUKlabourmarketexperiencehasbeennotablebecauseunemploymentrateshavenotrisenasfarandasfastas in previousperiods ofrecession oras in otherlargerEuropean economies,andthereforeledtosignificantunemploymentdurationdependency.Whathasbeennoticeablehasbeenthesustainedexperienceofstubbornly lowratesofwagegrowthacrosstheUK,butparticularlyinlaggingregions(Greggetal.,2014).Thismayhavehadfarmoreimpactonbothtippingthebalancebetweentherelativeattractivenessofself-employmentversuspaidemployment,andontheleveloflocaldemandfortheproductsandservicesofnewbusinessventures.Itreinforcesthesuggestionthatlocalunemploymentratesmaynotconveyfullinformationaboutlocal economic factors.
Onefurtherkeythemeintheliteratureisthedifferingself-employmentcontextformenandwomen(Brush,1992;Hughes,2003).IntheUKin2014themaleself-employmentratewas19%oftheeconomicallyactive,whereastherateforwomenwasonly10%.Howeverfemaleself-employmentgrewovertheprevioustenyearsby53%,comparedtoagrowthrateof16%formen(computedfromUKLabourForceSurveyestimates).Recentresearchalsostronglysuggeststhatmotivatingfactors,influencingchoice,aredifferentbetweenthetwo(Dawsonetal.,2014),althoughwhetherpeopleindividualsgiveobjectiveresponsestorecallquestionsaboutwhytheychoseself-employmentatsometimeinthepastis open to debate.
Tosummarisethediscussionsofar,thispaperrevisitsthequestionofwhetherthelikelihoodthatanindividualwillchoose(transitioninto)self-employmentishigherorlowerifthatindividuallivesinalocalitywhereunemploymentishigher.Thepaperalsoaddressesthesamequestionwithreferencetothelevelofwagesinthelocality.Importantly,inusingalarge-scalelongitudinalmicrodatasource,thisanalysisisconductedwhilstcontrollingforvariationinindividualskilllevelsandotherdemographicfactors.Inparticulartheanalysisfocusesonwhetherthelikelihoodofself-employmentisfurtheraffectedbywhethertheindividualwaspreviouslyunemployedorinactive,andfocusesoninvestigatingpotentialdifferencesbetween menandwomen.
3.Data and methodology
Theremainderofthepaperundertakesanempiricalinvestigationofself-employmentstatuschoicesandtransitionsusingrecentUKlongitudinaldataattheleveloftheindividual,drawnfromUnderstandingSociety(USoc),theUK’shouseholdlongitudinalsurvey.ThissurveybothextendsandembedstheformerBritishHouseholdPanelSurvey(BHPS).BHPS,whichwasconductedpriortotheestablishmentofUSocbetween1991to2008,providedacoresampleof5000households(withsomesubsequentregionalsampleboosts).USocwasinitiatedin2009withasampletargetof40,000householdsdrawnfromastratified,clusteredsampleofUKresidentialaddresses(includingamainstagesample,formerBHPShouseholdsandanewethnicminorityboostsample).Theachievedwave1sample,collectedoveratwo-yearperiod2009-2010was39,802households,comprising101,086individualsacrossallagesofwhom27,103wereinemploymentorself-employment.Samplewavescollectedareonanannualfrequencyacrosstwoyearoverlappingperiods,with4wavescurrentlyreleasedtoresearchers,i.e.2009-10,2010-11,2011-12and2012-13.Precisedatesofinterviewarerecorded.BuckandMcFall(2012)providefurthertechnicaldetailsofthesurveydesign.Infollowingahouseholddesign,theachievedWave1samplehascharacteristicsthatareverysimilar to the UK governmentLabour Force Survey.
Thereissomesampleattritionbetweenwavesduetolossofcontactorrefusaltoremainaparticipant.BetweenWaves1and2approximately20%ofthesamplewaslost(BuckandMcFall,2011),howeverattritionratesinsuccessivewavesweremuchlowerandinlinewithcomparablelongitudinalsurveys,internationally.Theinclusionofadditional“temporarysamplemembers”offsetsattrition;thesearerecruitediftheyjoinoriginallysampledhouseholds.
Table3providessummarydataonthescaleofself-employmentandtransitionsintoself-employmentinthesample.Theoverallrateofself-employmentrisesfrom13.3%to13.9%overthefouravailablesamplewaves.InlinewithofficialUKgovernmentdata(LabourForceSurvey)therateofself-employmentisconsiderablyhigherformenthanforwomen.Howevertherateofincreaseishigherforwomen.Notalltheself-employedarebusinessowners–someareabletoregisterwiththetaxauthoritiesasself-employedbecausetheyaresub-contractors,freelancingorworkinginsomeotherformofnon-businessownershipself-employment.Selfemployedbusinessowners(i.e.assoleownerorinpartnership)compriseoverthree-quartersoftheself-employed,invirtuallyidenticalproportionsformenandwomen,andshowingverysimilarratesofgrowthtooverallself-employment.
Thegrowthintheself-employmenttotalisreflectedinthegrowingnumbersoftransitionsbetweenwavesintoself-employment,bothfrompreviouspaidemploymentandfrompreviousinactivity(fromunemploymentorfromwithdrawalfromthelabourforceforotherreasons).Anumberoffeaturesstandout.Thefirstisthattheproportionofalltransitionsintobusinessownershipisjustbelow70%,andthereforelowerthantheproportionofbusinessownersintheself-employedstockasawhole.Inotherwordsthereappearstobesomegrowthinthescaleofsubcontractingandfreelancingactivitythatisnotconnectedtonewbusinessventuring.Thesecondisthataroundhalfofalltransitionsintoself-employmentarefrominactivityratherthanpaidemployment.Thissuggeststhatsignificantnumberschoseself-employmentasanalternativetoworklessness–howeverthiscouldreflectbothoreitherofa“necessity”motivearisingfromweaklocallabourmarketconditions and an “encouraged worker” effect arising from improving local economic
conditions.Thethirdfeatureisthatthegrowthintransitionsoverthefouryearsisstrongerforwomenthanformen,andtheproportionofwomenthattransitionfrominactivityishigherthanformen.Howeveritisimportanttoqualifytheseobservationsbynotingthattheabsolutenumbersoftransitionsobservedinwhatbeginsasaverylargesamplearequitesmall,andtherefore notnecessarilyafullyreliable guide to patterns in theoverallpopulation.
Inordertoaddressthequestionoftherelationshipbetweenself-employmentpatternsandlocaleconomicconditions,OfficeforNationalStatistics(ONS)dataon380localauthoritydistrictunemploymentratesandearningslevelsarelinkedintothemicrodatafiles.4Earningsaremeasuresatthelowerquartile(25thpercentile)level,inordertoreflectearningstowardstolowerpartofthedistributionasanappropriatecomparatorforthoseinparticularwhomaytransitionintoself-employmentfrominactivity.5Bivariatecorrelationanalysisofself-employmentstatusandtransitionswiththeselocaleconomicindicatorsisthenconducted.Self-employmentstatusandtransitionsintoself-employmentarethenmodelledinaprobabilisticframeworkusingbinarydependentvariableregression(probit)analysis.Thisallowsforcontrolforthemediatingimpactofdemographicfactorsincludingage,gender,ethnicity,educationandfamilybackground.Becausethedataarelongitudinal,withmultipleobservationsforeachindividual,itispossibletoperformthisanalysisbothusingconventionalregressionpoolingavailabledatafromthedifferentwaves,aswellasusingrandomeffectsmodelinginwhichvariationintime-invariantindividualcharacteristicsisincorporatedwithintheregressionerrorstructure,thoughtheassumptionthatthesuccessive
4Thisrequiredaccesstoinformationonthelocalauthoritydistrictofresidenceofeachsample household in each wave,informationthatis not released in thepublic accessdatafilesforUnderstandingSociety.TheauthorisgratefultotheUKDataServiceandtheUniversityof Essexfor grantingpermission to access this information.The ONSdataareextracted usingtheONSNomisWebservice,andrelatetoGreatBritain.NorthernIrelandisexcludedfromtheanalysisbecauseitsdataarecollectedonadifferentbasisandnotavailablethroughONS.5Infact it makes little difference to theresults ifthe localmedian level of earnings is used.
correlationoferrortermsforaparticularsampleindividualisconstantovertime(GuilkeyandMurphy, 1993; Arulampalam,1999).
The regression structure to model the probability for individual i at time t of
(transitioninto)self-employment,???,conditionalofasetofcovariatesxincludingthelocal unemployment rate and local earnings in thepreviousyear, takes thefollowingform:
Pr(??? ≠0|???) =Φ(????+??? )
(1)
where???=??+???inthecaseoftherandomeffectsestimation method, and??? =??? in
thecaseofthe“conventional”(pooled)probitmethod(althoughreportedestimatesinclude
standarderrorswhicharecorrectedforclusteringbymultipleobservationsoneachindividualsamplemember).Theexplanatorypoweroftherandomeffectsformversustheconventionalpooledformisusuallydescribedinthecalculationofρ,theproportionofthetotalerror
2
2?
,since?2 =1
variancecontributedbythepanellevelerrorvariance??,definedas?=?2+1?
byconstruction.ModelestimationwasperformedusingStataversion14.Because,inthe
caseoftherandomeffectsmodel,thelikelihoodfunctionisnotdefinedanalytically,theestimationmethoduses a Gauss-Hermite quadrature numericalapproximationmethod.
Alongsideindicatorsofconditionsinthelocaleconomy(unemployment,earnings)thevectorofcovariatesincludesarangeofotherfactorscommonlyusedinstudiesofself-employmentorentrepreneurshipchoice(DunnandHoltz-Eakin,2000;Taylor,2001;Georgellisetal.,2005;NiittykangasandTervo,2005;ColombierandMasclet,2008;Parker2009).Theseincludegender,age,levelofeducationalattainment,ethnicity,entrepreneurialparentageandrural/urbanlocation,aswellashighlevel(EUNUTS1)regionalindicatorsand time controls to capture anyaggregateeconomic cycle impact.
4.Findings
a)Correlationanalysis
Table4reportsbivariatecorrelationsandthesignificancelevelsoftwo-samplet-testscomparingmeanvaluesoflocalunemploymentratesandlocalmediangrossweeklyearningsfortheself-employed(transitioners)andothers.Resultsarereportedforall,andformenandwomenseparately.Allofthesearecorrelationsaresmall;howeverinmanycasesthedifferencesbetweenthegroupsarestatisticallysignificant.Thefirstrowofthetableshowsthatself-employmentisstatistically significantlylowerinlocalitiesofhigherunemployment.Thefindingholdsforbothmenandwomen.Forearningsthepictureismorecomplex.Althoughacrossthefullsampleself-employmentishigherinlocalitieswithhigher(lowerquartile)earnings,formenthereisaweakbutstatisticallysignificantnegativeassociation.Furtherdowninthefifthrowofthetable,similarcorrelationpatternsforself-employedbusinessownerarefound,particularforunemployment.However,self-employedbusinessownership is statisticallysignificantlylowerwherelower quartileearningsarehigher forbothmenandforwomen.Butacrossthecombinedsample,levelsofself-employmentwouldappeartobedepressedinlocalitieswherethelocallabourmarketisweaker,andencouragedwherelocalspendingpower(earnings)arehigher.Weaknegative,butsignificantseparatecorrelationsformenandwomenmaysuggestthatlowerlevelsofpay inthelowerpartofthepaidemploymentmarketdo raise the attractiveness of “goingit alone”.
ThesecondandsixthrowsoftheTable4reportassociationsbetweentransitionsintoself-employment/businessownershipandlocalunemploymentandearnings.Higherlocalunemployment is associated with slightly lower transitions; higher local earnings are
associatedwithhighertransitions.Theseassociations,althoughstatisticallysignificant,aresmallinsize.They suggestthatthenetimpactofimprovinglocaleconomicprosperity isoneof“prosperity-pull”.Anystrongeffectthatmightpushindividuals,whoarefinding itdifficult to findsatisfactorylocal employment, into choosingself-employment is absent in thedata.Thisisfurtherconfirmedbyseparatingtransitionsintothosethatarefrompaidemploymentandthosefrominactivity.Thereisnoindicationhereofastrong“push”effectfrominactivityintoself-employment.Formen,thenegativeassociationbetweenself-employmenttransitionsandlocalunemploymentdisappears.Forwomen,itisstronger.Women appear morelikelyto be attracted into self-employment iflocal conditions improve.
Finallythebottomrowofthetableexploitsinformationfromanadditionalquestionaskedof16to21yearolds,concerninglikelihoodofbeingself-employedinthefuture.6Thisvariableappearstocorrelatepositivelyandstatisticallysignificantlywithbothlocalunemploymentratesandlocalearningsforyoungmen.Anearningscorrelation,butnotanunemploymentcorrelation,isfoundforwomen.Thissuggeststhathigherunemploymentinthelocality may servetoencourageideasaboutbusinessstart-up.Onetheotherhandthesizeofthe correlationcoefficient is muchhigher forearningsandis significant foryoungmen andwomen,suggestingthatany“recession-push”motiveismorethanoffsetbythe“pull”effectcreatedbyavibrant locality.
6Thequestion(whatistheprobability,between0and100%,thatyouwillbeself-employedinthefuture)wasaskedinWaves2and3only.Respondentswereaskedto indicateperceivedprobability rangingfromzeroto100%.Themeanstatedprobabilitywas34%from8692 available responses.
b)Multivariate regressionanalysis
TheresultsinTable4areinconclusiveonwhetherlocalunemploymentratesandlocalearning levelscontributethesame or differentinformationaboutthelocalattractivenessofself-employment.Thisarguesinfavourofpayingattentiontotheresultsfrommultivariateanalysis.Table5reportsprobitregressionanalysisforself-employedstatus,whilsttable6reports a similar analysisfor transitions into self-employment.
InTable5threedifferentmodelspecificationsarereported.Columns1and2reportconventionalprobitmodels,althoughstandarderrorsareadjustedtoaccountforclusteringbecausethedatastructureentailsmultipleobservationsforthesameindividual.Column1includeshighlevelregionalbinaryvariablesandbinaryvariablesfortimeperiodofobservation,definedonsix-monthlyintervals.Column2excludesthese.Column3reportstheresultsofarandomeffectsprobitestimation,asdescribedinequation(1)above.Thehighestimatedvalueofρsuggestsstronglythattherandomeffectsformulationisthepreferredmodel.
Inallthreemodelsthereisasignificantnegativeassociationbetweenthelocalareaunemploymentrateandtheprobabilitythatanindividualwillbeinself-employment.Inthefirstmodel,wherethereareadditionalcontrolsfortimeandhigh-levelregions,thecoefficientsizeissmall,althoughstatisticallysignificant(theestimatedmarginaleffectissuchthata1percentagepointincreaseintheunemploymentrateisassociatedwithafallintheprobabilityofself-employmentof0.002).Suppressionoftimeandregionalcontrolsdoublesthissizeofthisassociation.Theseresultspointtoaneteffectfromlocalunemployment,whichisconsistentwithopportunityor “pull”effectsbeingstrongerthanany
“push” effect from unemployment into self-employment,evenafter controllingfortheimpactofvariationinlocalearningscapacity.Theassociationbetweenself-employmentandlocalearningslevelsisnotasconsistentintheestimates.Inthefirstcolumnlowerearningsareassociatedwithhigherself-employment,andtheeffectisstatisticallysignificant.Howeversuppressionoftimeandregionalcontrolscausetheassociationtochangesignandlosestatisticalsignificance.Incolumn3theassociationbetweenself-employmentandlocalearningsbecomesstronglypositiveandstatisticallysignificant.Thisagainisconsistentwithopportunityeffectsbeingstrongerthanany“push”effectsintoself-employmentfrompoorlocallabour market conditions.
Theothercontroleffectsrevealanumberofdemographicassociations.Consistentwiththedescriptivestatistics,womenaresignificantlylesslikelytobeself-employed.Olderindividualsaremorelikelytobeself-employed.Thereisamixedpatternofassociationswithethnicity:whitenon-British,AsianandChinese/EastAsiangroupstendtobemorelikelytochooseself-employment,otherthingsequal,comparedtowhiteBritishindividuals.Ontheother hand in theconventionalprobitequationsAfrican-Caribbeangroups are less likelyto beinself-employment.Self-employmentprobabilitiesarelowerifanindividual’shighestlevelofeducationalattainmentwasavocationalqualification,butothereducationaleffectsarenotsignificant.Havingafatherwhowasabusinessownerwhenyouwere14yearsofageissignificantlyassociatedwithahigherprobabilityofself-employment;althoughnosimilareffectisobservedformother’soccupationalstatus.Finallyindividualswhoareresidentinruralareasare, other things equal, significantlymorelikelyto bein self-employment.
InTable6aspreadofsevendifferentspecificationsarepresentedforyear-on-yeartransitionsintoself-employment,exploringboththedifferenceinmodelestimationmethod
andinteractionsofthelocallabourmarketvariableswithgenderandanindividual’sprioreconomicstatus(activityorinactivity).Columns1to3duplicatethoseinTable5.Onceagaintherelationshipwithlocalunemploymentisastatisticallysignificantnegativeone.Theindividualprobabilityoftransitionintoself-employmentfallsifthelocalunemploymentrateishigher.Marginaleffectsaresmall.Aone-percentagepointincreaseinunemploymentisassociatedwithadrop intheprobabilityoftransitionintoself-employmentofbetween0.0004and0.0005(fromitsmeanof0.014).Forlocalearnings,incolumns2and3withnohigher-levelregionalcontrols,theassociationwiththeprobabilityofself-employmenttransitionisapositiveandsignificantone.Again,althoughtheeffectisstatisticallysignificant,themarginaleffectsaresmall(0.002fora£100increaseinlowerquartileweeklyearningsincolumn3).Theemergingpictureisthesameinbothtables:evidenceforslightlyweakerbutstatisticallysignificantopportunityeffectsfromworseninglocaleconomicvibrancy.
Incolumns4and5unemploymentandearningsareinteractedwithgendertoinvestigatedifferencesintheseassociationsbetweenmenandwomen.Theoverallpictureisthatopportunityor“pull”associationsbetweenself-employmenttransitionsandlocallabourmarketconditionsaremuchstrongerforwomenthanformen.Thisappearstothecaseforbothlocalunemploymentratesandearningslevels.Intherandomeffectsmodelthefemaleself-employmenttransitionratemarginaleffectis0.0013foreachpercentagepointfallinunemployment, around three timesaslargeas thecombinedgender estimate.Forearnings themarginaleffect is twiceas large at 0.004.
Column6includesfurtherinteractioneffectstoinvestigatedifferencesbetweentheprobabilitiesoftransitionfromactivityorinactivity.Unemploymentandearningsgender
interactionsarefurtherinteractedwithabinaryvariableidentifyingthosewhowereinactiveinthepreviousyearofthesurvey.Thecoefficientsontheseadditionalinteractionsshouldthereforebeinterpretedastheadditionalimpactoflocallabourmarketconditionsforapreviouslyinactiveindividualcomparedtoonewhowasinpaidemployment.Forlocalunemploymenttheseinteractionsarenegativeandstatisticallysignificant,forbothmenandwomen.Theopportunity-damagingeffectofrisingunemploymentisstrongerforthoseconsideringself-employmentfrominactivitycomparedtothoseconsideringaswitchfrompaidemployment.Thesameeffectisalsofoundforlocalearnings.Lowerearningsreducetransitionprobabilitiesfurtherforthoseswitchingfrominactivitycomparedtothosefrompaidemployment.
TheestimatedvaluesofρfortherandomeffectsmodelsinTable6arearound0.1ineachcase.AlthoughmuchlowerthaninTable5,theseestimatesareallaroundtwiceaslargeastheirestimatedstandarderrors,andthereforestillshowingthatrandomeffectsformulationsare the preferred models.
Finallycolumn7inTable6includesinteractionsofthelocalunemploymentratewiththeindividual’sself-reportedlengthofeconomicinactivity (inyears),constructedfromUSocemploymenthistoryrecallquestionnaireschedules.7Inthecaseofthosewhoareinemploymentratherthaninactivethisnumberissettozero.ThepurposeofthisistoinvestigatethefindingrecentlyreportedbyAudretschetal.(2014)that it islong-termunemploymentratherthanunemploymentpersethathasanadverseimpactonnewfirmformation.Theresultsreportedareforapooledprobitestimationbecauseinthiscasethe
7Thesamplesizedropsbecausearound20%ofindividualsinthesamplehaveitemnon-response in the USocemployment historyschedules.
errorcomponentcorrelationisnotsignificantlydifferentfromzero.8Therearetwofeaturesoftheseresults.Thefirstisthattheinteractionwiththelengthoftimespentinactiveattractsstatisticallysignificantcoefficientestimatesforbothmenandwomen.Computedmarginaleffectsimplythatanadditionalyearofinactivitylowerstheprobabilityoftransitiontoself-employmentby0.02%(andbyslightlymoreformen).Thesecondfeatureisthatcontrollingforinactivitydurationincreasesthesizeandsignificance(toalevelof7%)ofthepositiveassociationbetweentheunemploymentrateandtheprobabilityoftransitionformen.Thissuggeststhattheremaybeaslightnet“push”effectintoself-employmentformenwhoareonlyrecentlyunemployed,supportingthefinding in Audretschetal. (2014).
TurningtotheothercovariatecoefficientestimatesinTable6, theseshowbothsimilaritiesanddifferencesindemographicsassociationswiththoseinTable5.Thesimilaritiesareas follows. Women areless likelyto transition into self-employment(amarginaleffectofaround0.01andthereforequantitativelysignificantgiventhesamplemeantransitionprobability of0.014).Thepatternofdifferencesacrossethnicgroupsbroadly holdsacrossthepatternofcoefficientsfromstatustotransitionprobabilities,withWhitenon-BritishandOtherethnicgroupssignificantlymorelikelytotransitionintoself-employment.Othercoefficientestimates,althoughcarryingthesamesignsasinTable5,arenotstatisticallysignificant.Havingabusinessownerfatherincreasesthetransitionprobabilityandthoseinruralareasaremorelikelytotransitiontoself-employment.Thekeydifferencesareforageandeducation.Olderindividualsaresignificantlylesslikelytotransitionintoself-employmenteventhoughtheyaremorelikelytobefoundinself-employment.Educationaleffectsarerelativetothereferencegroupofthosewithbelowage16ornoeducationalqualifications. Both university/college graduates and those with age 16 school leaving
8Consequentlytheunreportedrandomeffectscoefficientestimatesareveryclosetothosereported.
qualificationsaremorelikelytotransitionintoself-employment.Groupswithintermediatelevelqualificationsarenotlesslikely.Thecoefficientforvocationalqualifications,whichwasnegativeandsignificant in Table 5, usuallyremains negative in themodels for transitionsbutisnotsignificant.Theseresultsareconsistentwithself-employmentbeingattractivetoa)thosewithbothprofessionaloccupationalskillsandb)thosewhoventurebusinessactivitythatdoesnotrequirehighlevelsofformaleducationalattainmentinpursuitofreturnshigherthan those available in the paidlabour market.
5.Discussion
Themainfindingtoemergefromthisanalysisisthatlocaleconomicandlabourmarketconditionsappeartoexertalargelypositiveinfluenceonthelikelihoodofchoosingself-employment.Inotherwords,improvinglocalunemploymentandearnings(inthelowerpartoftheearningsdistribution)arepositivelyassociatedwiththeprobabilitythatanindividualwillchooseself-employment(Table5)orchoosetotransitionintoself-employment(Table6).Thereislittleornosuggestionofanynet“recession-push”effectonself-employment.Itwasnotedearlierthattheextantcross-sectionalliteraturehasprovedtobeagnosticonthekeyquestionofapositiveornegativeassociationbetweenself-employmentandunemployment(Parker2009).However,thispaperfinds,evenduringthedifficulteconomic conditions in the immediate aftermath ofthe2007-8globalfinancialcrisis,ratherstrongerevidenceforalocaldemand“pull”effect.Thisstandsinstarkcontrasttotheweightofevidencein time-seriesanalysesforrecession-pusheffects (Parker 2009).Ifthereisany“push”influenceofrisingunemploymentorfallingpaidemploymentearningsintoself-employment,formostindividualsitismorethanoffsetbytheopportunity-damagingeffectsthatrisinglocalunemploymentorfallingwageshaveontheattractivenessofalocalityasa
placeinwhichtodobusiness.Thescaleoftheassociationsisstrongerforthosewhoareconsideringtransitionintoself-employmentfromeconomicinactivity,andsuggeststhatentrepreneuriallabourforceparticipationdecisionsareinfluencedbytheexpectedreturnsfrombusinessventuring.Women,inparticular,seemtobeencouragedtoswitchfromlabourforcewithdrawal into business venturing activitybyimprovements in thelocal economy.
Theonlyevidencefoundforthe“push”hypothesisintheanalysisisformen,particularlythoseconsideringtransitioningfrom(poorly)paidemploymentorfromarelativelyshortperiodofinactivity,orunemployment.Thisevidenceisstatisticallyquiteweak.However,thelongersomeonehasbeenunemployedandthehigherthelocalunemploymentrate,thenthelesslikelythatpersonistotransitionintoself-employment,offeringfurther supportfor Audretschetal. (2014), from a differenttype ofdata source.
Insummary,itisdifficulttoconcludefromthisanalysisthatthereisreallymuchevidencefora“push”effectofrisingunemploymentintoself-employmentandsmallscalebusinessventuring.Itcannotberuledoutthatboth“push”and“pull”effectsarelargebutoffseteachothertotheextentthatthenetassociationissmall.A“push”effectmayexistinthemindsofthoseconsideringself-employment.Howeveranysucheffectisverysignificantlyoffsetbythenegative“opportunity”effect.Higherlocalunemploymentandlowerlocalearningslevelssignalthatthepotentialgainsfrombusinessventuringhaveworsened.Alternativelyfallingunemploymentandrisingwagesencouragebusinessventuring,ratherthanencouragetheself-employedtoswitchoutofself-employmentintobetter payingandlessriskywagedemployment.
CurrentUKpolicyinthisareaisfocusedontheprovisionthe“newenterpriseallowance”whichprovidesupto26weeksofincomesupportaswellasaccesstosomestart-uploancapital,targetedinparticularattheinactiveandunemployed.9Whilstpoliciesofthisnaturemaysupporttheunemployedtotakeadvantageofentrepreneurialopportunities,theparadoxintheresultshereisthatthismayhavemoreimpactinleadingregionsratherthaninlaggingones,wherelocalunemploymentratesarelowerandearningsatthelowerendofthedistributionarehigher.Inturnthismay,atthemargin,worsenratherthanlessenregionalinequalities.
Althoughtheseresultsfailtoidentifyanynetrecession-pusheffectforself-employmenttransitions,inTable4apositivecorrelationbetweenlocalunemploymentandyoungmen’sreportedlikelihoodoffutureself-employmentwasreported.Localunemploymentmaystimulateinterestinself-employmentevenifactualtransitionsdonotoccur, particularlyformen.This is investigatedfurtherinasimilarregressionanalysisreportedinTable7.Theseresultsdosuggestaweakly significantpositiveassociationbetweenthelocalunemploymentrateandtheprobabilitythata16to21yearoldmaleattachestofutureself-employment.Foryoungwomenanylocallabourmarketeffectseemsto focusmore on opportunity,asindicatedbythepositive earnings coefficient.10
Thepresentstudyhasalsoattemptedtohighlightthepotentialvalueoflarge-scalelongitudinalsurveydataforresearchonentrepreneurshipandself-employment.Inparticulartheanalysisfocusesontransitionsintoself-employment,ratherthantheself-employment
9See
10TwootherfeaturesstandoutinTable7.Thefirstisthatinterestinself-employment,evenwithinasmallagerangesample,fallswithage.ThesecondisthatyoungAfrican-Caribbeanandotherethnicgroupshavehigherinterestinself-employment,eventhoughforthefirstofthese there is no evidenceof associationwith higher transitionrates in Table6.
status.Thisisimportantbecausetheformerarefarmorelikelytobeinfluencedbyrecentmovementsinlocaleconomicconditions,whereasthelattermayreflectindividualdecisionsmadeinitiallyatatimeofverydifferenteconomiccircumstancesandsubsequentlyaffectedbyinertiaandaccumulatedexperience.Neverthelesslarge-scalequantitativeanalysisisnotwithoutitslimitations–inparticularitlackstheability toaddressinamorenuancedmannertheunderlyingcausalprocessesthatmayleadanindividualtoreflectonlocaleconomiccircumstanceswhenassessingtherangeofspecificopportunitiesandchoiceswhichthatindividual mayfaceat agiven point in time andplace.
6.Conclusions
Structuralgrowthinself-employmentintheUKbeganwellbeforetheonsetoftheglobalfinancialcrisis,buthascontinuedatasignificantpacesince,accountingforaverysignificantproportion ofrecentUK jobsgrowth.In the termsofthequestion posed in thetitletothispaper,theanalysispresentedinthispapersuggests,despitesomesignificantrecentcommentary,thatthisistheoutcomeofvoluntarychoicesmadebyindividualswhoinpartrespondtolocalsignalsabouteconomicopportunity.Theself-employeddonotappeartobereluctantconvertstoentrepreneurship,“encouraged”intobusinessstart-upactivitybytheabsenceof(wellpaid)localjobs.Thispaperhasinvestigatedthestrengthofanyassociationoflocalunemploymentratesandlocallowerquartileearningswithself-employmentstatusandthelikelihoodofindividualtransitionintoself-employment.Associations,whileinsomecases,statisticallysignificant,arenotquantitativelystrong.Opportunity-pulleffectsappeartobestrongerforwomenthanformen.Neverthelessthesefindingsdooffersomeinsightintowhyself-employmentgrowthhasbeenstrongestinregionswhereunemploymentislower,
earningsarehigherandthereforeeconomicconditionsforentrepreneurialopportunitymorefavourable.Thecontinuedgrowthinself-employmentappearstobestructuralratherthancyclicalinthesensethatitislikelytobedrivenby longertermsmovementsinperceptionsofentrepreneurshipand availabilityofresources and skills tosupportthosechoosingself-employedbusinessownershipasacareermove.Policiestosupporttheunemployedintoentrepreneurshipmayexacerbateratherthanalleviateregionalinequalities,becausethey mayhavemoreimpactinfastgrowinglocalitiesratherthanlagging ones.Inlagging ones,start-upsupportfortheunemployedandinactivemayoffercoldcomfortiflocaldemandconditionsremainweak.Regionalpolicytostimulatedemandfortheoutputsofnewbusinesses,ratherthanencouragesupplyofentrepreneuriallabour,islikelytobemuchmoreeffectiveinraisingentrepreneurialactivity.
References
AmitR.(1994),“‘Push’and‘pull’entrepreneurship”,inBygraveW.D.,BirleyS.,Churchill
N. C., Gatewood E., Hoy F., Keeley R. H. and Wetzel, Jr W. E. (Eds) Frontiers inEntrepreneurshipResearch1994.Proceedingsof the14thAnnualEntrepreneurshipResearchConference,BabsonCollege,Wellesley, MA(available at:
Armington,C.andAcs,Z.(2002),“Thedeterminantsofregionalvariationinnewfirmformation”,Regional Studies 36: 33-45.
Arumlampalam,W.(1999),“Anoteontheestimateeffectsinrandomeffectsprobitmodels”,
OxfordBulletin of Economics and Statistics, 61: 597-602.
Audretsch,D.B.,Dohse,D.andNiebuhr,A.(2014),“Regionalunemploymentstructureandnew firmformation”,KielWorking Paper No. 1924, Kiel:Institute for theWorld Economy.
Baumgartner,H.J.andCaliendo,M.(2008),“Turningunemploymentintoself-employment:effectivenessoftwostart-upprogrammes”,OxfordBulletinofEconomicsandStatistics,70(3):347-373.
Biehl,A.M.,Gurley-Calvez,T.andHill,B.(2014),“Self-employmentofolderAmericans:do recessionsmatter?”,SmallBusinessEconomics, 42: 297-309.
Benjamin,D.andKochin,L.(1979),“SearchingforanexplanationofunemploymentinInterwar Britain”,Journal of PoliticalEconomy,87(3):441-478.
Brush,C.G.(1992),“Researchonwomenbusinessowners:pasttrends,anewperspectiveandfuture directions”,EntrepreneurshipTheoryand Practice,16(4):5-30
Buck,N.andMcFall,S.(2012),“UnderstandingSociety:designoverview”,Longitudinaland LifeCourse Studies,3(1);5-17.
Caliendo,M.andKunn,S.(2014),“Regionaleffectheterogeneityofstart-upsubsidiesforthe unemployed”,Regional Studies,48(6):1108-1134.
Centre for Cities(2015),CitiesOutlook 2015,London:Centre for Cities(January).
Clark,KandDrinkwater,S(2000),“Pushedoutorpulledin?Self-employmentamongethnicminorities in England and Wales”,Labour Economics, 7: 603-628.
Colombier,N.andMasclet,D.(2008),“Intergenerationalcorrelationinself-employment:some furtherevidence from FrenchECHPdata”,SmallBusinessEconomics, 30: 423-437.
D’Arcy,C.andGardiner,K.(2014),JusttheJob–OraWorkingCompromise?TheChanging Natureof Self-Employment in the UK,London:ResolutionFoundation(May).
Dawson,C.,Henley,A.andLatreille,P.(2014),“Individualmotivesforchoosingself-employment in the UK: does region matter?”RegionalStudies48(5):804-822.
Dunn,T.andHoltz-Eakin,D.(2000),“Financialcapital,humancapitalandthetransitiontoself-employment:evidencefromintergenerationallinks”,JournalofLaborEconomics,18:282-305.
Evans, D.S. and Leighton, L.S. (1990), “Small business formation by unemployed andemployedworkers”,Small BusinessEconomics,2(4):319-30.
Frisch, M. (1993), “Regional differences in new firm formation: evidence from WestGermany”,RegionalStudies,26(3):233-241.
Frisch,M.andStorey,D.J,(2014),“Entrepreneurshipinaregionalcontext:historicalroots,recentdevelopmentsandfuture challenges”,Regional Studies,48(6):939-954.
Georgellis,Y.,Sessions,J.G.andTsitsianis,N.(2005),“Windfalls,wealthandthetransitionto self-employment”,Small BusinessEconomics, 25: 407-25.
GiladB.andLevineP.(1986),“Abehavioralmodelofentrepreneurialsupply”,JournalofSmallBusinessManagement24(4), 45–53.
Gregg,P.,Machin,S.andFernández-Salgado,M.(2014),“Thesqueezeonrealwages–andwhatmight it take to endit”,National Institute Economic Review: 228(1): R3-R16.
Guilkey,D.K.,andJ.L.Murphy(1993),“Estimationandtesting intherandomeffectsprobitmodel”,Journal of Econometrics, 59: 301–317.
Hatfield,I. (2015), Self-Employment in Europe,London:Institute for PublicPolicyResearch.
HMRevenueandCustoms(HMRC)(2013),OnshoreEmploymentIntermediaries:FalseSelf-Employment:ConsultationDocumentURL: re_employment_intermediaries_-_false_self_employment.pdf (accessed16-1-15).
Henley,A.(2005),“Jobcreationbytheself-employed:therolesofentrepreneurialandfinancialcapital”,SmallBusinessEconomics, 25:175-196.
Hessels,J.,vanGelderen,M.andThurik,A.R.(2008),“Entrepreneurialaspirations,motivationsandtheir drivers”,SmallBusinessEconomics,31(3):323-39.
Hinz,T.andJungbauer-Ganz,M.(1999),“Startingabusinessafterunemployment:characteristicsandchancesofsuccess(empiricalevidencefromaregionalGermanlabourmarket,Entrepreneurship and RegionalDevelopment, 11: 317-333.
Hughes,K.(2003),“Pushedorpulled?Women’sentryintoself-employmentandsmallbusinessownership”,Gender,Work and Organization, 10: 433-454.
Hurst,E.andLusardi,A.(2004),“Liquidityconstraints,householdwealthandentrepreneurship”,Journal of PoliticalEconomy, 112: 319-47.
Lindh,T.andOhlsson,H.(1996),“Self-employmentandwindfallgains:evidencefromtheSwedishlottery”,Economic Journal, 106: 1515-26.
McManus,S.(2012),“Jobsatisfaction:anindicationofwhatweexpect,notjusttheworkwedo”,NatCenSocialResearch what-we-expect,-not-just-the-work-we-do(accessed15-May-2015)
Millán,J.M.,Congregado,E.andRomán,C.(2012),“Determinantsofself-employmentsurvival in Europe”,Small BusinessEconomics,38(2):231-258.
Niittykangas,H.andTervo,H.(2005),“Spatialvariationsinintergenerationaltransmissionof self-employment”,Regional Studies, 39: 319-32.
Parker,S.(2009),TheEconomicsofEntrepreneurship,Cambridge:CambridgeUniversityPress.
Parker,S.andRobson,M.T.(2004),“Explaininginternationalvariationsinself-employment:evidencefrom apanel ofOECD countries”,Southern EconomicJournal, 71: 287-301.
Reynolds, P., Storey, D. and Westhead, P. (1994), “Cross national comparisons of thevariation in new firm formation rates”,RegionalStudies, 28: 443-456.
Ritsila,J.andTervo,H.(2002),“Effectsofunemploymentonnewfirmformation:micro-levelpaneldataevidencefromFinland”,Small BusinessEconomics,19(1):31-40.
Robson,M.T.(1998),“Self-employmentinUKregions”,AppliedEconomics,30(3):313-322.
Storey,D.J.(1991),“Thebirthofnewfirms–doesunemploymentmatter?Areviewoftheevidence”,Small Business Economics, 3: 167-178.
Taylor, M.P. (2001), “Self-employment and windfall gains: evidence from panel data”,
Economica, 68: 539-65.
Tervo, H. (2006), “Regional unemployment, self-employment and family background”,
AppliedEconomics, 38(9): 1055-62.
Thurik,A.R.,Carree,M,vanStel,AandAudretsch,D.(2008),“Doesself-employmentreduce unemployment”,Journal of BusinessVenturing,23(6):673-686.
Table1: UKEmployment and Self-Employment Growth2008-2014
TotalEmployment‘000s / Employees‘000s(% of total) / Self-Employment‘000s
(% of total) / BusinessStock+
‘000s
2008 quarter 1 / 29,510 / 25,428
(86.2%) / 3,858
(13.8%) / 4,075
2014 quarter 2 / 30,430 / 25,630
(84.2%) / 4,551
(15.8%) / 4,367
Change / +920 / +202 / +693 / +292
Inflow 2009-2014* / 1,669
(36.5%)
Outflow 2009-2014* / 886
(23.4%)
Source:UK Officefor NationalStatistics
Notes:*inflowmeasuredaspercentageof2014totalself-employedwhohadenteredwithinlast5years;outflowmeasuredaspercentageof2009totalwhohadleftself-employmentcomparedto5yearspreviously.+registeredandunregistered(forVATandPAYE),0to10employees.
Table2: Changes in Self-Employment by UK NUTS1 Region,2008-2013
2008annualaverageself-employment‘000s / 2013annualaverageself-employment‘000s / Change‘000s / Change%England:
NorthEast / 112 / 125 / +13 / +11.6
North West / 363 / 426 / +63 / +17.4
YorksHumberside / 290 / 312 / +22 / +7.6
EastMidlands / 239 / 258 / +19 / +7.9
West Midlands / 288 / 337 / +49 / +17.0
East of England / 395 / 442 / +47 / +11.9
London / 571 / 679 / +108 / +18.9
South East / 577 / 676 / +99 / +17.2
South West / 383 / 424 / +41 / +10.7
Wales / 174 / 192 / +18 / +10.3
Scotland / 268 / 286 / +18 / +6.7
NorthernIreland / 123 / 112 / -11 / -8.9
Source:UK Office forNationalStatisticsandNorthernIreland,Deptof Trade,Enterprise andInvestment
Table3: USoc Sample Self-employment Levelsand TransitionRates
Wave12009-10 / Wave2
2010-11 / Wave3
2011-12 / Wave4
2012-13
Males and females:
Self-employment
(as %ofallemployment) / 3758
(13.3%) / 4046
(13.2%) / 3857
(13.8%) / 3729
(13.9%)
Self-employedbusinessownership(as %ofallemployment) / 2875
(10.2%) / 3138
(10.2%) / 2988
(10.7%) / 2899
(10.8%)
Alltransitions into self-employment / - / 477 / 585 / 569
Alltransitions into self-employedbusinessownership / - / 320 / 379 / 395
Transitionsfrominactivityintoself-employment / - / 251 / 297 / 291
Transitionsfrominactivityintoself-employedbusinessownership / - / 164 / 188 / 194
Malesonly:
Self-employment
(as %ofallemployment) / 2580
(18.4%) / 2802
(18.3%) / 2605
(18.7%) / 2513
(18.8%)
Self-employedbusinessownership(as %ofallemployment) / 1976
(14.1%) / 2164
(14.1%) / 2047
(14.7%) / 1937
(14.5%)
Alltransitions into self-employment / - / 309 / 346 / 340
Alltransitions into self-employedbusinessownership / - / 207 / 225 / 225
Transitionsfrominactivityintoself-employment / - / 154 / 171 / 154
Transitionsfrominactivityintoself-employedbusinessownership / - / 98 / 111 / 96
Femalesonly:
Self-employment
(as %ofallemployment) / 1178
(8.3%) / 1244
(8.1%) / 1252
(8.9%) / 1216
(9.1%)
Self-employedbusinessownership(as %ofallemployment) / 899
(6.3%) / 974
(6.3%) / 941
(6.7%) / 962
(7.2%)
Alltransitions into self-employment / - / 168 / 239 / 229
Alltransitions into self-employedbusinessownership / - / 133 / 154 / 170
Transitionsfrominactivityintoself-employment / - / 97 / 126 / 137
Transitionsfrominactivityintoself-employedbusinessownership / - / 66 / 77 / 98
Table4: Bivariate correlation analysis ofself-employmenttransitions and locallabourmarket conditions
Correlation(Two-sample t-test) / UR - all / Q25E-
all / UR -
males / Q25E -
males / UR -
females / Q25E -
females
Self-employmentstatus / -0.050
(0.000) / 0.076
(0.000) / -0.047
(0.000) / -0.013
(0.019) / -0.059
(0.000) / 0.009
(0.091)
Self-employmenttransition in previousyear from anystatus / -0.009
(0.003) / 0.035
(0.000) / 0.002
(0.692) / 0.015
(0.001) / -0.021
(0.000) / 0.017
(0.000)
Self-employmenttransition in previousyear fromemployment / -0.003
(0.519) / 0.037
(0.000) / 0.003
(0.605) / 0.005
(0.423) / -0.012
(0.046) / 0.018
(0.002)
Self-employmenttransition in previousyear from inactivity / -0.014
(0.001) / 0.031
(0.000) / -0.001
(0.846) / 0.021
(0.001) / -0.025
(0.000) / 0.018
(0.002)
Self-employedbusinessownershipstatus / -0.049
(0.000) / 0.048
(0.000) / -0.047
(0.000) / -0.032
(0.000) / -0.057
(0.000) / -0.014
(0.009)
Self-employedbusinessownershiptransition in previousyear from anystatus / -0.011
(0.001) / 0.022
(0.000) / 0.001
(0.886) / 0.008
(0.108) / -0.023
(0.000) / 0.004
(0.319)
Self-employedbusinessownershiptransition in previousyear fromemployment / -0.009
(0.054) / 0.024
(0.000) / 0.002
(0.755) / 0.001
(0.869) / -0.023
(0.000) / 0.004
(0.950)
Self-employedbusinessownershiptransition in previousyear from inactivity / -0.017
(0.000) / 0.026
(0.000) / -0.006
(0.438) / 0.014
(0.052) / -0.026
(0.000) / 0.006
(0.288)
Would liketo be self-employed in thefuture (age 16-21) / 0.022
(0.123) / 0.114
(0.000) / 0.053
(0.011) / 0.061
(0.003) / -0.004
(0.827) / 0.063
(0.001)
Source:author’scomputationsfromUnderstandingSocietyWaves 1 to 4
Notes:UR–localauthoritydistrictofresidenceunemploymentrate12monthspreviously;Q25E–localauthoritydistrict1stquartileweeklyearningsinpreviousyear;italicdenotesp-value below 0.1, bold italicbelow 0.05
Table5: Multivariate regression(probit)modelestimates for self-employment status
Coefficient(p-value) / (1) / (2) / (3)Probit / Probit / RandomEffectsProbit
Localunemploymentrate(lagged) / -0.012
(0.034) / -0.021
(0.000) / -0.054
(0.001)
Local1stquartileearnings(lagged,genderspecific
£’00s)(medianearnings incolumn 3) / -0.055
(0.023) / 0.031
(0.124) / 0.107
(0.010)
Gender(female=1) / -0.476
(0.000) / -0.423
(0.000) / -1.005
(0.000)
Age(years) / 0.020
(0.000) / 0.020
(0.000) / 0.053
(0.000)
Ethnicity(reference:white British)
White,non-British / 0.242
(0.000) / 0.270
(0.000) / 0.788
(0.000)
African-Caribbean / -0.183
(0.002) / -0.107
(0.059) / -0.197
(0.248)
Asian / 0.120
(0.005) / 0.160
(0.000) / 0.235
(0.075)
ChineseandotherEastAsian / 0.104
(0.209) / 0.158
(0.053) / 0.495
(0.039)
Other / 0.105
(0.194) / 0.145
(0.075) / 0.399
(0.095)
Highesteducationalattainment(reference:belowage 16 schoolqualifications)
Universityorcollegefirstdegree orhigher / 0.014
(0.567) / 0.031
(0.207) / -0.008
(0.923)
VocationalqualificationincludingHNDs / -0.250
(0.000) / -0.239
(0.001) / -0.679
(0.002)
A-levels orequivalentaged 18 / -0.065
(0.075) / -0.053
(0.141) / -0.162
(0.152)
O-levels/GCSEsorequivalentaged 16 / -0.035
(0.172) / -0.022
(0.389) / -0.088
(0.278)
Fatherwasbusinessowner/employer / 0.125
(0.008) / 0.128
(0.006) / 0.313
(0.027)
Motherwasbusinessowner/employer / 0.003
(0.962) / 0.004
(0.944) / 0.021
(0.908)
Rurallocation / 0.215
(0.000) / 0.193
(0.000) / 0.462
(0.000)
NUTS 1 regionalcontrols / Yes / No / No
Time controls / Yes / No / Yes
N / 31337 / 31337 / 31337
NT / 66083 / 66083 / 66083
LogL / -24537.7 / -24618.2 / -14540.0
PseudoR-sqrd / 0.069 / 0.066 / -
Proportionoferrorvariancecontributed bypanellevelerrorvariance, ρ / - / - / 0.992
Source:author’scomputationsfromUnderstandingSocietyWaves 1 to 4
Notes:Sample–alleconomicallyactive.P-values(columns1and2)arecomputedafteradjustmentofstandarderrorsforclustering byindividual.Italicdenotessignificanceat0.1orhigher,bolditalicat0.05orhigher.Medianearningsareusedincolumntoachievemodellikelihoodconvergence.
Table6: Multivariate regression(probit)modelestimates for transitionstoself-employment
Coefficient(p-value) / (1) / (2) / (3) / (4) / (5) / (6) / (7)Probit / Probit / Randomeffectsprobit / Probit / Randomeffectsprobit / Randomeffectsprobit / Probit
Localunemploymentrate(lagged) / -0.012
(0.052) / -0.015
(0.005) / -0.016
(0.003)
Localunemploymentx male / 0.003
(0.646) / 0.003
(0.684) / 0.012
(0.132) / 0.014
(0.073)
Localunemploymentx female / -0.043
(0.000) / -0.045
(0.000) / -0.027
(0.015) / -0.024
(0.019)
Localunemploymentx male x previouslyinactive / -0.033
(0.022)
Localunemploymentx female xpreviouslyinactive / -0.047
(0.005)
Localunemploymentx male x yearsdurationofpreviousinactivity / -0.006
(0.000)
Localunemploymentx female x yearsdurationofpreviousinactivity / -0.005
(0.000)
Local1stquartileearnings(lagged,genderspecific£’00s) / -0.013
(0.620) / 0.069
(0.001) / 0.067
(0.002)
Local1stquartileearnings xmale / 0.046
(0.058) / 0.045
(0.086) / 0.029
(0.286) / 0.017
(0.559)
Local1stquartileearnings xfemale / 0.125
(0.000) / 0.123
(0.001) / 0.069
(0.070) / 0.095
(0.029)
Local1stquartileearnings xmale x previouslyinactive / 0.064
(0.001) / 0.074
(0.000)
Local1stquartileearnings xfemale x previouslyinactive / 0.124
(0.000) / (0.105)
(0.000)
Gender(female=1) / -0.263
(0.000) / -0.212
(0.000) / -0.226
(0.000) / -0.260
(0.058) / -0.262
(0.086) / -0.241
(0.116) / -0.343
(0.055)
Age(years) / -0.007
(0.000) / -0.007
(0.000) / -0.007
(0.000) / -0.007
(0.000) / -0.007
(0.000) / -0.008
(0.000) / -0.008
(0.000)
Ethnicity(reference:white British)
White,non-British / 0.154
(0.008) / 0.183
(0.002) / 0.195
(0.001) / 0.176
(0.003) / 0.188
(0.002) / 0.195
(0.001) / 0.164
(0.011)
African-Caribbean / -0.033
(0.568) / 0.029
(0.605) / 0.033
(0.571) / 0.029
(0.610) / 0.033
(0.572) / 0.028
(0.632) / -0.035
(0.612)
Asian / -0.033
(0.460) / 0.006
(0.899) / 0.007
(0.881) / 0.003
(0.951) / 0.004
(0.934) / -0.011
(0.815) / -0.014
(0.795)
ChineseandotherEastAsian / 0.093
(0.260) / 0.137
(0.092) / 0.150
(0.080) / 0.133
(0.102) / 0.145
(0.089) / 0.129
(0.137) / 0.137
(0.149)
Other / 0.131
(0.080) / 0.168
(0.024) / 0.180
(0.025) / 0.165
(0.027) / 0.177
(0.027) / 0.172
(0.034) / 0.145
(0.099)
Highesteducationalattainment(reference:belowage 16 schoolqualifications)
Universityorcollegefirstdegreeorhigher / 0.180
(0.000) / 0.196
(0.000) / 0.204
(0.000) / 0.196
(0.000) / 0.203
(0.000) / 0.235
(0.000) / 0.144
(0.000)
VocationalqualificationincludingHNDs / -0.015
(0.839) / -0.005
(0.951) / -0.006
(0.941) / -0.008
(0.911) / -0.009
(0.898) / 0.003
(0.970) / 0.008
(0.921)
A-levels orequivalentaged 18 / 0.005
(0.899) / 0.021
(0.606) / 0.016
(0.695) / 0.022
(0.591) / 0.017
(0.681) / 0.025
(0.552) / 0.017
(0.721)
O-levels/GCSEsorequivalentaged 16 / 0.061
(0.033) / 0.074
(0.010) / 0.074
(0.013) / 0.075
(0.009) / 0.075
(0.012) / 0.086
(0.004) / 0.071
(0.033)
Fatherwasbusinessowner/employer / 0.097
(0.035) / 0.099
(0.032) / 0.108
(0.028) / 0.098
(0.034) / 0.106
(0.031) / 0.102
(0.040) / 0.078
(0.142)
Motherwasbusinessowner/employer / -0.028
(0.676) / -0.025
(0.713) / -0.028
(0.690) / -0.022
(0.748) / -0.025
(0.723) / -0.019
(0.795) / 0.006
(0.929)
Rurallocation / 0.124
(0.000) / 0.101
(0.000) / 0.106
(0.000) / 0.104
(0.000) / 0.108
(0.000) / 0.110
(0.000) / 0.073
(0.011)
NUTS 1 regionalcontrols / Yes / No / No / No / No / No / No
Time controls / Yes / No / Yes / No / Yes / No / No
N / 49553 / 49553 / 49553 / 49553 / 49553 / 49553 / 39046
NT / 106523 / 106523 / 106523 / 106523 / 106523 / 106523 / 86224
LogL / -7857.8 / -7885.0 / -7875.4 / -7873.4 / -7864.1 / -7847.8 / -5789.4
PseudoR-sqrd / 0.031 / 0.028 / - / 0.029 / - / - / 0.035
Proportionoferrorvariancecontributed bypanellevelerrorvariance, ρ / - / - / 0.086 / - / 0.084 / 0.101 / -
Source:author’scomputationsfromUnderstandingSocietyWaves 1 to 4
Notes:Sample–alladults.P-values(columns1,2and4)arecomputedafteradjustmentofstandarderrorsforclusteringbyindividual.Italic
denotessignificance at 0.1 or higher,bold italicat 0.05 or higher.
Table7: Multivariate regressionmodelestimates for future self-employment likelihood(16-21year olds)
Coefficient(p-value)Localunemploymentx male / 0.005
(0.078)
Localunemploymentx female / -0.004
(0.178)
Local1stquartileearnings xmale / 0.019
(0.118)
Local1stquartileearnings xfemale / 0.027
(0.038)
Gender(female=1) / -0.037
(0.560)
Age(years) / -0.012
(0.000)
Ethnicity(reference:white British)
White,non-British / 0.069
(0.108)
African-Caribbean / 0.118
(0.000)
Asian / 0.016
(0.326)
ChineseandotherEastAsian / 0.009
(0.817)
Other / 0.165
(0.000)
Fatherwasbusinessowner/employer / 0.070
(0.007)
Motherwasbusinessowner/employer / 0.002
(0.946)
Rurallocation / 0.022
(0.078)
NUTS 1 regionalcontrols / No
Time controls / No
N / 3670
NT / 4939
R-sqrd / 0.041
Source:author’scomputationsfromUnderstandingSocietyWaves 1 to 4