Discussion PaPer series

IZA DP No. 10689

Joel BlitMikal SkuterudJue Zhang

Immigration and Innovation: Evidence from Canadian Cities

APrIl 2017

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA Guiding Principles of Research Integrity.The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our time. Our key objective is to build bridges between academic research, policymakers and society.IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.

Schaumburg-Lippe-Straße 5–953113 Bonn, Germany

Phone: +49-228-3894-0Email: publications@iza.org www.iza.org

IZA – Institute of Labor Economics

Discussion PaPer series

IZA DP No. 10689

Immigration and Innovation: Evidence from Canadian Cities

APrIl 2017

Joel BlitUniversity of Waterloo

Mikal SkuterudUniversity of Waterloo and IZA

Jue ZhangUniversity of Waterloo

AbstrAct

APrIl 2017IZA DP No. 10689

Immigration and Innovation: Evidence from Canadian Cities*

We examine the effect of changes in skilled-immigrant population shares in 98 Canadian

cities between 1981 and 2006 on per capita patents. The Canadian case is of interest

because its ‘points system’ for selecting immigrants is viewed as a model of skilled

immigration policy. Our estimates suggest unambiguously smaller beneficial impacts of

increasing the university-educated immigrant population share than comparable U.S.

estimates, whereas our estimates of the contribution of Canadian-born university graduates

are virtually identical in magnitude to the U.S. estimates. The modest contribution of

Canadian immigrants to innovation is, in large part, explained by the low employment rates

of Canadian STEM-educated immigrants in STEM jobs. Our results point to the value of

providing employers with a role in the immigrant screening process.

JEL Classification: J61, J18, O31

Keywords: immigration, innovation, immigration policy

Corresponding author:Mikal SkuterudDepartment of EconomicsUniversity of Waterloo200 University Avenue WestWaterloo, ON, N2L 3G1Canada

E-mail: skuterud@uwaterloo.ca

* We would like to thank David Green, Isaac Ehrlich, Jenny Hunt, Bill Kerr, Bill Lincoln, Garnett Picot, Chris Worswick, and seminar participants at the University of Melbourne, University of British Columbia, University at Buffalo, University of Waterloo, Immigration, Refugees, and Citizenship Canada (IRCC), and the 2015 and 2016 Canadian Economics Association Meetings for valuable comments. The authors acknowledge financial support from the Social Science and Humanities Research Council of Canada (SSHRC).

2

1 Introduction

Animportantconsequenceoftheeconomicturmoilbroughtaboutbythefinancialcrisisof2008

wasadecreaseinvoters’supportofimmigration.Thisdevelopment,whichhasbeenparticularlyevident

intheU.S.andtheU.K.,hasputincreasingpressureonpro-immigrationpoliticianstojustifytheeconomic

benefitsofcontinued large-scale immigration.Todoso, increasing referencehasbeenmade inpolicy

discussions to the burgeoning economics literature exploring the `wider’ benefits of immigration,

includingeffectson internationaltradeflows,entrepreneurship,and,perhapsmostsignificantly,given

the growing consensusof its importance to long-termeconomic growth, on innovation.Although the

precisetheoreticalmechanismsthroughwhichdiversityincreasesinnovationarelesswelldeveloped,the

empirical literatureprovides remarkablyconsistentevidenceof theproductivity-enhancingbenefitsof

increasingethnicdiversitywithinworkplaces,cities,andcountries.1

Forgovernmentpolicymakersresponsibleforimmigration,thecriticalquestionishowtoharness

thisgrowth-enhancingpotentialofethnicdiversity.Inthisrespect,theeconomicsliteraturelinkingskilled

immigration with higher patenting rates is arguably not only the most relevant, but also the most

compelling.BeginningwithU.S.studiesbyPeri(2007),Chellaraj,Maskus,andMattoo(2008),Huntand

Gauthier-Loiselle (2010), and Kerr and Lincoln (2010), but now also including a number of European

studies(Bosetti,Cattaneo,andVerdolini(2012);Ozgen,Nijkamp,andPoot(2012),Parrotta,Pozzoli,and

Pytlikova(2014),Nathan(2014a)),thisliteraturehasattractedconsiderableattentioninthepolicyworld.

The results from these studies consistently suggest that increasing skilled immigration, particularly of

immigrantseducatedinscience,technology,engineering,andmathematics(STEM)fields,hasasignificant

positive impacton thenumbersofpatents that are created. For example,Hunt andGauthier-Loiselle

(2010) findthataonepercentage-point increase in theshareofastate’spopulationwhoarecollege-

educatedimmigrantscanbeexpectedtoincreasestate-levelpatentspercapitaby9-18%.Comparingthe

magnitudeofthiseffecttowhatisimpliedbythedifferentialpatentingrateofimmigrantsobservedin

individual-leveldata,theyconcludethatanimportantpartofthiseffectreflectsapositiveexternalityof

immigrants on the patenting rates of native-born Americans. The potential of immigrants to raise

1Thenotionof`widereffects’ofimmigrationisduetoNathan(2014b).Theliteraturelinkingethnicdiversityand

innovationisinterdisciplinarywithpapersinpsychology(VanKnippenberg,DeDreu,andHoman2004),sociology

(Herring2009),managementstudies(ElyandThomas2001;Richard,McMillan,Chadwick,andDwyer2003),and

economics.

3

innovationlevelsnotonlydirectlythroughtheirownpatents,butalsobymakingnativesmoreinnovative,

makesastrongeconomiccaseforimmigration.

In this paper, we examine the Canadian case in order to inform the innovation-enhancing

potential of immigration in a setting inwhich a `points system’ is used to screen skilled immigrants.

Canada’s `points system’ is seen bymany as amodel of effective skilled immigration policy; theU.K.

adoptedapointssystemin2008anditisregularlypointedtoasanoptioninongoingU.S.immigration

reformdiscussions.TheCanadiancaseisalsoimportantbecauseCanadaconsistentlyranksamongthe

world’slargestimmigrant-receivingcountriesmeasuredasaproportionofitspopulation.Betweenthe

mid-1980sandmid-1990s,bothCanada’sannualinflowofnewpermanentresidentsandtheshareofthe

inflowadmittedunderthe`pointssystem’morethandoubled.Consequently,theshareoftheCanadian

working-agepopulationcomprisedofuniversity-educatedimmigrantsincreasedfrom2.1%intheearly

1980sto3.3%intheearly1990sand6.4%bythemid-2000s.

GivenCanada’ssuccessatattractingskilledimmigrants,thereisthepotentialforexceptionally

large effects of immigration on innovation in the Canadian case. However, there is also substantial

evidencepointingtosignificant labourmarketchallengesofCanadianuniversity-educated immigrants,

whichsuggestthatthe labourmarketskillsofCanadian immigrantshavenotkeptpacewiththe large

increaseintheireducationlevels(ClarkeandSkuterud2013,2016;Clark,Ferrer,andSkuterud2017).Itis

anopenquestionwhether thepoor earningsperformanceof Canadian immigrants, possibly resulting

fromthecrudenessofthecriteriausedbythe`pointssystem’toscreenhumancapital,ismirroredintheir

contributionstoinnovation.Inparticular,whiletheCanadian`pointssystem’givesconsiderableweight

toforeignsourcesofeducationandworkexperience,there isevidencethatforeignsourcesofhuman

capitalaredevaluedbyCanadianemployers(GreenandWorswick2012;SkuterudandSu2012).

TheprimarychallengeinexaminingtheCanadacaseisitsrelativelysmallpopulation,whichlimits

the geographic variation in immigrant population shares. Nonetheless, relating changes in university-

educated immigrantshareswithin98Canadiancitiesbetween1981and2006tochanges inpatenting

rates, we obtain estimates that are unambiguously smaller than those found by Hunt and Gauthier-

Loiselle (2010) (hereafterHGL) usingU.S. data. This remains true evenwhenwe restrict attention to

university-educatedimmigrantswhowereeducatedinaSTEMfield.Ontheotherhand,theestimated

effectofCanadian-bornuniversitygraduatesonpatentingratesisvirtuallyidenticalinmagnitudetothe

HGLestimateforU.S.natives,suggestingthatthesmallermagnitudeofourimmigrantestimatesdoesnot

reflect greater measurement error in our data or something intrinsic to the Canadian economy or

4

innovation sectors. Overall, our analysis suggests that increasing the university-educated immigrant

populationshareinCanadamayhavecontributedtoraisingpatentingrates,butonlymodestly,andany

spillovereffectsofimmigrantsonnativepatentingarelikelyminimal.

AnimportantpolicyquestionistowhatextenttheweakercontributionofCanadianimmigrants

toinnovationthatweidentifyisrelatedtothebroaderlabourmarketchallengesofCanadianimmigrants

identifiedelsewhere. Indeed,whenwe isolate theeffectofuniversity-educated immigrantswhowere

educatedinaSTEMfieldandarecurrentlyemployedinaSTEMoccupation,ourestimatesbecomemuch

largerandstatisticallysignificant.TherelativelysmallCanadianestimatesthereforeappearto, in large

part,reflecttherelativelylowemploymentratesofCanadianimmigrantsinSTEMjobs,includingamong

thoseeducatedinSTEMfields.WhileweprovidenodirectevidenceonwhyCanadianSTEM-educated

immigrantsfacegreateremploymentbarriersthantheirU.S.counterparts,thedifferenceisconsistent

withU.S.immigrantsbeingrelativelypositivelyselectedowingtoagreaterroleofemployersinimmigrant

selectionandhighereconomicreturnstoskillinU.S.labourmarkets.

The remainder of the paper is organized as follows. In the following section, we discuss the

relevanceoftheCanadiancontext.Insection3,wedescribeourmethodologicalapproach,includingthe

datathatweemploy.InSection4wediscussourresultsindetail.Inthefinalsection,wesummarizeour

mainfindingsanddiscusstheirpolicyrelevance.

2 TheCanadiancontext

TheCanadianImmigrationActof1962endedthehistoricalpracticeofselectingimmigrantsonthebasis

oftheircountryoforiginandreplaceditoverthefollowingdecadewitha pointssystem’thatemphasized

the human capital of migrants. The success of the Canadian `points system’ in raising the average

educationlevelsofitsimmigrantpopulationhasledanumberofcountries,includingAustraliaandthe

U.K.,tofollowitsapproach,andhasreceivedmuchattentioninrecentimmigrationreformdiscussionsin

theUnitedStates.ThekeyrationaleunderlyingtheCanadianapproachisthathumancapitalisastronger

predictor of long-run economic success than the extent towhich an immigrant’s skillsmatch current

labourmarketneeds.Moreover,currentlocallabourmarketneedsaredifficulttoidentifyempiricallyand,

areoftenshort-lived,and theapproach is inpractice impractical, since immigrantsare free tochoose

wheretheysettle.However,withinCanadatherehasbeengrowingcriticismofthisapproachinresponse

toevidenceofadeteriorationintheabilityofCanada’sskilledimmigrantstoobtainjobscommensurate

5

with their levels of education andexperienceobtained abroad (seePicot and Sweetman (2012) for a

reviewofthisliterature).2

ThelevelofinnovationinCanadahashistoricallybeenlowerthanthatoftheUnitedStates.The

economyinvestsasmallerfractionofGDPonresearchanddevelopment(2.0%inCanadaversus2.5%in

theU.S.in2006)andgeneratesfewerpatentspercapita(19.9patentsper100,000inCanadaversus48.0

patentsper100,000intheU.S. in2006).Prevailingexplanationsforthisgapincludedifferencesinthe

industrialmix(inparticular,Canada’shistoricalrelianceonnaturalresources),ahigherdegreeofforeign

ownershipinCanada,andtherelativelysmallersizeofCanadianfirms.However,thetwocountriesdonot

differinthefractionoftheirworkforcesemployedinSTEM.AsreportedbyBecksteadandGellatly(2006),

theshareofemploymentinscience,engineering,andrelatedoccupationswas,forCanadaandtheU.S.

respectively,9.8%and9.6%in1981/80,11.7%and11.3%in1991/90,and13.6%forbothin2001/00.

Given the lower level of patenting activity in Canada,wemight expect lower patenting rates

amongCanadianskilledimmigrantsandthattheygeneratelesspatentingspilloversonnatives.However,

thefocusofouranalysisiswhetherCanada’s pointssystem’forscreeningskilledimmigrants,inparticular

onthebasisoftheireducationalattainmentlevels,hasresultedinCanadianimmigrationhavingalarger

proportionalimpactonpatentingrates.Toprovidesomeinitialsenseofthemagnitudesofthesechanges,

inFigure1weplotbothnational-levelpatentspercapitainCanadaandtheU.S.between1980and2006

andthesharesoftheirpopulationsaged25andovercomprisedofuniversity-educatedimmigrants. In

bothcountries,theuniversity-educatedimmigrantshare increasedconsistentlyovertheentireperiod.

GiventheCanadiansystem’semphasisonskilledimmigration,theCanadiansharein1980wasmorethan

twicetheU.S.share(2%comparedto0.7%).Overthefollowing25years,Canadacontinuedtoattract

moreskilledimmigrantsasafractionofitspopulation,sothatbythemid-2000snearly6.4%ofitsworking-

ageCanadianpopulationwereuniversity-educatedimmigrants,comparedto4.2%intheUnitedStates.

GiventheevidenceinHGL,thisincreaseshouldhaveservedtoraisepatentingratesproportionally

moreinCanadathanintheUnitedStates.Interestingly,theCanadianpatentingratedid,infact,increase

2ThishasledtheCanadiangovernmenttomakesignificantpolicyshiftsinrecentyearstowardsgivingemployersa

greaterroleinimmigrantselection.Inparticular,asufficientconditionforobtaininganinvitationforpermanent

residencyunderthenewExpressEntrysystemforprocessingapplications,introducedinJanuary2015,isajob

offerfromaCanadianemployer.Joboffersforforeignworkersmust,however,clearalabourmarkettestintended

toensurethattheemployerwasunabletofillthejobdomestically.

6

moreoverthisperiodthantheU.S.rate.3Whereaspatentspercapita(x100,000)nearlytripledinCanada

(fromabout6.9in1980to19.9in2006),theyonlydoubledintheU.S.(25.9in1980to48.0in2006).Of

course,theincreaseinpatentingratesimpliedbyeventheupperboundestimateofHGL(an18logpoint

increaseinpatentspercapitafroma1percentage-pointincreaseintheuniversity-educatedimmigrant

share)aremuchsmallerthanthelogpointincreasesthateitherCanadaortheU.S.actuallyexperienced.

Ofcourse,therearemanyotherfactorsservingtoraisepatentingratesbesidesimmigration.Moreover,

thesenational-levelcorrelationscouldbeentirelymisleading.Toplausiblyidentifythecausalimpactof

Canada’sskilledimmigrationonitspatentingrate,weneedastrategytoisolateasourceofincreasesin

skilledimmigrantpopulationsharesthatareplausiblyindependentof increasesinpatentingratesthat

wouldhaveoccurredevenintheabsenceofanychangesinskilledimmigrantpopulationshares.

3 Methodology

WefocusoncomparisonstotheresultsofHGLforthreereasons.First,theirresultsarethemost

general, as they are focused on college-educated shares in the overall population, as opposed to

internationalstudentsorH-1Bvisaholders.Thismakesitpossibletoconductmoredirectcomparisons.

Second,HGLhasattractedthemostinterest.4Third,theyfindevidenceoflargedirectandspillovereffects

of immigrants on U.S. patenting rates.5 However, rather than examine state-level (or province-level)

immigrationshares,asHGLdo,werelateimmigrantsharestopatentratesatthecitylevel.6Specifically,

weconstructa1981-2006balancedpanelofCanadianCensusMetropolitanandAgglomerationAreas

(CMA/CAs)withobservationsonskilledimmigrantpopulationsharesin98citiesevery5years.7Ourcities

rangeinpopulation(age15-70)in2006fromalowof8,448toahighof3,684,821,with66citiesabove

25,000individuals,46above50,000,26above100,000,and7above500,000.

3Bothcountriesexhibitupwardtrendingpatentingratesuptothedot-combubbleburstingin2001.FortheU.S.,in

particular,thisincreasewasfollowedbyalargedecline,whichmayhavebeendue,inpart,toadropinthesuccess

rateofpatentapplicationsattheUSPTO,particularlyinthe“drugsandmedicalinstruments”and“computersand

communications”fields(Carley,Hedge,andMarco2003).Itisimportanttonotethat,becausewehavecollected

patentsgranteduptoNovember2014,andthatamongpatentsgrantedin2013only1.8%ofthemtook8yearsor

longertobegrantedfromthedateofapplication(whichweuseinthefigure),datatruncationlikelyexplainsonlya

smallfractionofthisdecrease.4CitationcountsforHGLinGoogleScholarare417and56inWebofScienceasofMay2016.Incomparison,the

secondmostcitedpaper,KerrandLincoln(2010),has291and48citations,respectively.5KerrandLincoln(2010)donotfindstrongevidenceofspillovereffects.

6Withonly10Canadianprovinces,twoofwhichaccountforroughly60%ofthenationalpopulation,ananalysisat

theprovincelevelisnotviable.7ACMAisdefinedasoneormoreadjacentmunicipalitiescenteredonapopulationcorewithatleast100,000.ACA

musthaveacorepopulationofatleast10,000.

7

Weestimatetheskilledimmigrantsharesofthepopulationusingthemasterfilesofthe1981,

1986, 1991, 1996, 2001, and 2006 Canadian Censuses, which provide 20% random samples of the

Canadianpopulation.Skilledimmigrantsaredefinedinfouralternativeways:(i)university-educated;(ii)

university-educatedinaSTEMfield;(iii)university-educatedandemployedinaSTEMoccupation;or(iv)

university-educatedinaSTEMfieldandemployedinaSTEMoccupation.Theappendixprovidesdetails

on howwedefine STEM fields of study andoccupations in the various Census years. In addition,we

distinguishbetweenSTEM-educatedimmigrantswithCanadianandforeigndegrees,whichweestimate

usinginformationonyearsofschoolingandageatimmigration.8Incaseswherethepopulationshares

aredefinedusingfieldofstudy,welosethefirstyearofdatainourpanelbecausefieldofstudywasnot

identifiedinthe1981Census.

Skilled immigrant population shares in Census years are related to the number of patent

applications(percapita)withincitiesoverthefollowing5years.Thefive-yearlagisnotonlyconvenient

for maximizing our sample size using the quinquennial Canadian Censuses, but is also justified by a

separate analysis we conducted suggesting that the impact of changes in the composition of the

populationonpatentapplicationcountspeaksfouryearsafterthechange.9Weconstructpatentcounts

atthelevelofthecityandyearusingUnitedStatesPatentandTrademarkOffice(USPTO)dataonpatents

grantedtoinventorsresidinginCanada.Alternatively,wecouldhaveexaminedpatentsgrantedbythe

CanadianIntellectualPropertyOffice(CIPO)toCanadianinventors.However,thiswouldhaveresultedin

usobservingonlyasmallsubsetofpatentedCanadianinventions,sinceCanadianinventorstendtopatent

intheU.S.andforegopatentinginCanadaaltogether,duetothemuchlargersizeoftheU.S.market.10

PatentsareassignedtocitiesbylinkingtheaddressofinventorstoCanadianCMA/CAs.Where

patents containedmultiple inventors, we assigned fractions of patents to cities, so that each patent

8Specifically,weassumeschoolingisstrictlycontinuous,sothatyearsofschoolingplus6identifiestheageof

schoolcompletion.Comparingthisagetotheageatimmigrationidentifieswhethertheterminaldegreewas

obtainedinCanadaorabroad.Theresultingvariablecontainssomemeasurementerrorwhereschoolingisnot

continuousandwhereinternationalstudentsobtainCanadianschoolingpriortolanding.SkuterudandSu(2012)

showthattheconsequencesofthismeasurementerrorarenegligibleinestimatingearningstoforeignand

Canadianschooling.9Werelatedchangesinacity’spopulationfromagivenethnicitywithchangesinthenumberoffuturepatent

applicationsbymembersofthatethnicityresidinginthatcity.WethankBillKerrforgenerouslyprovidinguswith

dataonthepredictedethnicityofpatentinventorsbasedontheirnames(seeKerrandLincoln2010).10WeconductedaseparatesearchonthewebsitesoftheCIPOandtheUSPTOforpatentsfiledintheyear2000

withatleastoneCanadianinventorandfound1,136CIPOand5,195USPTOpatentsmeetingthecriteria.To

furthertestthepremisethatCIPOpatentsarelargelyasubsetofUSPTOpatents,wemanuallysearchedtheUSPTO

databaseforthefirst100Canadian-inventorCIPOpatentsappliedforin2000andfound93unambiguousUSPTO

matchesand2additionalprobableones.Thesedataareavailablefromtheauthorsuponrequest.

8

receivedequalweight.Forexample,apatentwithtwoinventorsfromTorontoandonefromKitchener-

Waterlooiscountedastwo-thirdsofapatentforTorontoandone-thirdforKitchener-Waterloo.Patents

areassignedayearbasedontheapplicationdateofthepatent(notthegrantdate),sincethiscoincides

mostcloselytotheactualdatethattheinnovationtookplace.Becauseweonlyobservepatentsgranted

uptoNovember2014,ourpatentcountsforthefive-yearwindowfollowing2006(theyears2007-2011)

will be lower due to data truncation.However, among patents granted in 2013,we find that 58%of

patentsweregrantedwithin3yearsofapplication,75%within4years,86%within5years,93%within6

years,and96%within7years.Ourestimatedpatentcountswill,therefore,beroughly18%lowerinthis

windowthantheyshould,butthisvariationshouldbeabsorbedinthe2006yearfixedeffect.

Ourbaselineempiricalmodelestimatesaspecificationascloseaspossibletothefirst-difference

(FD)weightedleastsquares(WLS)specificationofHGL.Wethenextendthisspecification,byincludinga

richer set of controls intended to address the possible endogeneity of within-city changes in skilled

immigrantpopulationshares.Specifically,weestimatetheequation:

5

1

( )( ) ( )

log( ) ( ) ( )

( ) (1981) ( ) ( )

cj c c

m nc c c

c c c

patents t jsm t sn t

pop t pop t pop t

X t Z y t t

b b

q ed

=

æ ö+ç ÷ æ ö æ öç ÷D = D + D +ç ÷ ç ÷ç ÷ è ø è øç ÷è ø

D + + +

å (1)

wherepatentsc(t+j)isthetotalnumberofpatentsgrantedtoinventorsresidingincitycthatwerefiledin

yeart+j;popc(t)isthepopulationaged15andover;smc(t)and snc(t) arethenumberofskilledimmigrants

andnatives(age15andover),respectively;Xc(t)isavectoroftime-varyingcontrolvariables;Zc(1981) is

avectorofcontrolsmeasuredin1981,intendedtocapturetheinfluenceofinitialconditions;y(t)isaset

ofCensusyearfixedeffects;εc(t)isarandomerrorpotentiallycorrelatedacrossyearswithincities;andΔ

is the first-difference between Census years. The parameter βm identifies the proportional effect of

increasingtheskilledimmigrantpopulationsharebyonepercentagepointonpatentspercapita,both

directlyandthroughpossiblespilloversonthepatentsofnatives.

FollowingHGL,webeginbyestimatingequation(1)includinglogmeanageinXc(t)andbothlog

mean income and log population in Zc(1981). We then extend the model by adding to Xc(t): (i) the

employmentrateand(ii)theexpectednumberoflogpatentspercapitabasedonthedistributionofa

9

city’spatentsbetween1972-1980acrosspatentclassesandthenational-levelnumberofpatentswithin

thosepatentclassesacrossCensusyears.This lattercontrolvariable,whichweborrowfromKerrand

Lincoln(2010), is intendedtocapturespuriouscorrelationsbetweenhistoricalsectoraldistributionsof

innovationacrosscitiesandsubsequentimmigrationflows.Intheextendedversionofthemodel,wealso

include a set of region-year fixed effects,where regions include theMaritimes,Quebec,Ontario, the

Prairies,andBritishColumbia.Finally,weallowthelogmeanincomecontrolvariabletovaryacrossCensus

years.Giventheconsiderablevariationincitysizesinoursampleof98Canadiancities,thevarianceof

the error term across city observations will vary considerably. To improve the efficiency of the FD

estimatorwethereforeweightalltheregressionsbycitypopulationsize.11

Itis,ofcourse,possibletoestimateequation(1)usingafixed-effects(FE)estimatorinstead.With

more than two time periods, the FE estimator produces different estimates than the FD estimator,

althoughbothestimatorsareconsistentunderthestrictexogeneityassumptionthattheright-hand-side

variables in equation (1) are uncorrelated with εc(t) across all Census years. Obtaining substantially

differentpointestimatesusingFE,thatisnotduetosamplingerror,providesevidenceagainstthestrict

exogeneityassumption.WehaveestimatedallthespecificationswereportusingaFEestimatorandnone

ofourmainfindingsaresubstantivelyaltered.

Thekeychallengeinidentifyingthecausalimpactofimmigrationonpatentsusinganarea-level

analysis is that we would expect skilled migration flows to be higher to cities that are experiencing

relativelylargeincreasesininnovationactivityforreasonsthatareentirelyindependentofimmigration.

Forexample,skilledimmigrationintheU.S.isdriveninlargepartbytherecruitingactivitiesofemployers,

throughtheH-1Bvisaprogram.Ifunobservedtechnologyshockssimultaneouslyleadtoincreasesinboth

patentsandthedemandforH-1Bworkers,theestimatesofβmwilltendtobeupwardbiasedestimates

ofthecausalimpactofimmigrants.Employerlabourdemandhas,however,historicallyplayedlittlerole

intheCanadian`pointssystem’,whichisusedtoscreenthevastmajorityofeconomicclassapplicants.

Moreover,thesystemhashistoricallybeencharacterizedbysignificantprocessingbottlenecks,makingit

11Specifically,weweightthefirst-differencedobservationsby(popc(t+1)-1+popc(t)-1)-1.AconcernwiththeWLS

approachistheinfluenceofTorontoontheestimates,givenitsrelativelylargepopulation.Thisisalsoaconcernin

theIVestimationdescribedbelow,inwhichtheinstrumentsarebasedonhistoricaldistributionsofimmigrants

acrosscities.ToassureourselvesthatourfindingsarenotdrivenbytheTorontoobservationalone,wehavealso

estimatedallourmodelsexcludingToronto.AlthoughthesenaïveFD-WLSestimatesdosuggestsomewhatlarger

beneficialimpactsofuniversityeducatedimmigration,thesearestillunambiguouslysmallerthanthoseinHGL(see

TableA1intheappendix),andourIVestimatesarealmostidenticaltothosereportedinTable5.Alternatively,we

havealsoestimatedunweightedregressionsforthelargest53cities(thosewithapopulationofatleast40,000in

1981).Theestimatesarealsolarger(seetableA2intheappendix)butstillsignificantlysmallerthanthoseinHGL.

10

arguablylesslikelythatsupply-drivenchangesinimmigrationflowstoCanadiancitiesarecorrelatedwith

latentcity-levelchangesinpatentingactivity.Nonetheless,eveninCanada,immigrantsultimatelydecide

in which city they will reside. To the extent that skilled immigrants choose to settle in cities where

increasesinpatentingratesarealreadyhappening,thereisstillreasontobeconcernedthattheresults

fromthenaïveestimatesofequation(1)areupwardbiased.

Acommonsolutiontothisinferenceproblem,initiallyproposedbyCard(2001),istoisolatethe

supply-pushcomponentofimmigrationflowstoaparticularcityusingattributesofcitiesthatareplausibly

unrelated to latent innovation trends.Thestandardapproach,whichwe follow, is to instrument local

skilled immigrantpopulationsusingpredicted immigrantpopulationsbasedon thehistorical city-level

settlement patterns of immigrants from particular origin countries and national-level populations of

immigrantsfromthosecountries.Thatis,weinstrumenttheskilledimmigrantsharesmc(t)inequation(1)

usingtheconstructedvariable:

(1976( ) ( ))c cj jj

sm t sm tl=å (2)

whereλcj(1976)istheshareof1976Canadianimmigrantsbornincountryjlivingincitycandsmj(t)isthe

national-level population of skilled immigrants from country j living in Canada in year t.12 Using first-

differencesoftheskilledimmigrantshares,theintuitionbehindtheinstrumentalvariables(IV)strategyis

that, for example, if the increase in the skilled immigrant population originating from Germany is

exceptionallyhighatthenationallevelbetweentwoCensusyears,wewouldexpectthecityofKitchener-

Waterloo(KW)toreceiveadisproportionatelylargeshareofthisincrease,notbecausetheseimmigrants

wereattractedby theexpectationofheightened innovativeactivity inKW,butbecause thehistorical

populationofGermanmigrantsresidinginKWandtheassociatedculturalamenitiestheyofferattracts

them.

12Toobtain1976immigrantcitypopulationsbyorigincountryweusedmobilityinformationinthepreviousfive

yearscontainedinthe1981Census,butrestrictedthesampletoimmigrantswholandedin1976orearlier.Wedid

not,however,restrictthesampletoskilledimmigrants,sinceculturalamenitiesthatattractimmigrantsarelikely

tobesharedacrosseducationgroups.Wealsogroupedcountriesintoregionswithsharedcultures,inorderto

reducemeasurementerrorintheestimatesofλcj(1976).ThegroupsaretheCaribbeanandBermuda(Frenchand

non-Frenchareseparategroups),CentralAmerica,SouthAmerica(Frenchandnon-French),Germany,France,

WesternEurope(excludingGermanyandFrance),EasternEurope,Scandinavia,SouthernEurope,Australia/New

Zealand/U.K.andcolonies,Sub-SaharanAfrica(Frenchandnon-French),otherAfrica(Frenchandnon-French),

Oceania(Frenchandnon-French),WesternAsiaandMiddleEast,India/Bangladesh/Pakistan,China/Hong-

Kong/Taiwan,Singapore/Malaysia/Indonesia,Korea,SouthAsia(excludingIndia,Pakistan,andBangladesh),and

restoftheworld.

11

4 Results

Beforeexaminingtheresultsofourregressionanalysis,inTable1wereportsamplemeansofthevariables

usedintheregressionsseparatelybyCensusyear.Themeansareweightedbycitypopulations,sothat

theyarerepresentativeoftheCanadianpopulationresidingwithinoneofCanada’slargest98cities.Note

that thepatent rates inTable1are roughly five times larger than those inFigure1because theyare

cumulativesumsofpatentsinthe5yearsfollowingtheCensusyear(thedependentvariableinequation

1).Consistentwiththenational-levelCanadianpatentingrateinFigure1,thefirstrowofTable1indicates

thataveragepatentingratesinCanada’scitiesincreasedconsistentlybetweentheearly1980sand2000s,

resultinginanearthreefoldincrease.Thequestionis,towhatextentdidskilledimmigrationcontribute

tothisincrease?

InthefollowingrowsofTable1,wereportskilledpopulationsharesseparatelyforimmigrants

andnatives.TheoverallimmigrantsharewithinCanada’slargestcitiesincreasedby4.6percentagepoints

between 1981 and 2006, which is larger than the change in the national-level share, reflecting the

increasing concentrationof new immigrants inCanada’s three largest cities – Toronto,Montreal, and

Vancouver. More important, all of this increase appears to be accounted for by university-educated

immigrants,astheirsharealoneincreasedby5percentagepoints(from2.7%to7.6%).Giventhatthe

Canadian `points system’hasneverdiscriminatedon thebasisof fieldof study, it ispossible that this

increase isaccountedforprimarilyby immigrantswhowereeducatedandemployed insectorswhere

patentingactivityisrare.Inthatcase,theireffectonpatentratesmayhavebeenmuchsmallerthanthe

HGL estimateswould predict. However, not only did the STEM-university-educated share increase by

about2percentagepointsbetween1986and2006,accountingforclosetohalfoftheoverallincreasein

the university-educated share, but by the early 2000s the share of university-educated Canadian

immigrantswhowere educated in a STEM field exceeded the comparable share forU.S. immigrants.

DefiningSTEMfieldsofstudysimilarlyusingtheU.S.NationalSurveyofCollegeGraduates(NSCG),33.6%

ofU.S.college-educatedimmigrantsin2003wereeducatedinaSTEMfield,comparedto37.4%and38.7%

ofCanadianuniversity-educatedimmigrantsin2001and2006,respectively.TheCanadian`pointssystem’

appears, therefore, to have been successful in not only raising the education levels of Canada’s

immigrants,butalsoinselectingimmigrantseducatedinSTEMfields.

Nonetheless, the Canadian research on the labour market performance of new immigrants

revealssignificantjob-educationmismatch.Foreign-trainedengineersdrivingtaxisismorethanacliché

inCanada(Xu2012).Giventhatthevastmajorityofpatentinghappensthroughcorporateresearchand

12

developmentactivities,challengesofSTEM-educatedimmigrantsinobtainingjobsinSTEMoccupations

mayhavelimitedtheimpactofSTEM-educatedimmigrantsonCanadianpatenting.Thereis,infact,some

evidence of this possibility in Table 1, as the population share comprised of university-educated

immigrantsfromSTEMfields increasedby2percentagepointsbetween1986and2006,buttheshare

alsoemployedinaSTEMoccupationincreasedbylessthan1percentagepoint.

In Table 2, we examine this education-job mismatch more closely by reporting conditional

probabilitiesof employment in a STEMoccupation separately for immigrants andnatives. The results

revealthatnotonlyareCanadianimmigrantsmorelikelytoholdauniversitydegreethantheirnative-

borncounterparts,butthisadvantagehasgrownsignificantlyovertime.Moreover,university-educated

immigrantsinCanadahavealwaysbeenmorelikelytobeeducatedinaSTEMfieldthantheirnative-born

counterpartsandthisdifferencehasalsobecome largerover time.By2006,nearly4-in-10university-

educatedCanadianimmigrantsweretrainedinaSTEMfield,comparedto2-in-10natives.However,the

probabilityofaSTEM-university-educatedimmigrantbeingemployedinaSTEMoccupationhastended

todecreaseovertime,whereasithasincreasedfornatives.Consequently,by2006therewasnearlya5

percentagepointgap in theSTEM-employment rateofCanadianSTEM-educated immigrants (0.37 for

natives,comparedto0.32forimmigrants).Incomparison,datafromtheNSCGindicatethatone-halfof

STEM-educatedimmigrantsintheU.S.wereemployedinSTEMjobsinboth1993and2003.Incontrast,

thecomparable rate forCanadianand theU.S.natives is similar (roughly0.4 inbothcountries).13We

wouldclearlyexpectthisshortfallintheSTEM-employment-ratesofCanadianimmigrantstohavelimited,

inasignificantway,thepotentialofCanada’sgrowingSTEM-university-educatedimmigrantpopulation

toboostCanadianinnovation.

A possible explanation for the low STEM-employment rates of STEM-educated Canadian

immigrantsisthatforeignsourcesofeducation,whichtheCanadian`pointssystem’valueshighly,may

resultinbarrierstoemployment,perhapsbecausethequalityofschoolingisloweronaverageorbecause

employers have more difficulty evaluating foreign credentials. Distinguishing between immigrants

educatedinCanadianandforeignuniversitiesprovidessomelimitedsupportforthispossibility.Rows6

and 7 of Table 2 show that the probability of being employed in a STEM job among STEM-educated

immigrantswithCanadiandegreeshasconsistentlybeenabout3percentagepointshigherthanforSTEM-

13AlthoughthefieldofstudyandoccupationclassificationsystemsinourCensusdataandtheNSCGaredifferent,

thefactthattheestimatedSTEM-employment-rateofSTEM-educatednativesaresimilarsuggeststousthatthe

muchloweremploymentrateofCanadianSTEM-educatedimmigrantsisnotbeingdriveninhowSTEMfieldsand

occupationsarebeingclassifiedinthetwodatasourcesorbyadifferentindustrialmixacrossthetwocountries.

13

educatedimmigrantswithforeigndegrees(theonlyexceptionbeingtheendofthedotcombubblein

2001,whentherateswereidentical).However,theimpactofthisemploymentgaphasbecomemagnified

astheshareofSTEM-university-educatedimmigrantswhograduatedfromaforeignuniversityincreased

fromabout50% in1986to57%in2006,presumablyreflectingthegrowing importanceof the `points

system’inimmigrantselection.Onceagain,wewouldexpectthistrendtohavelimitedthepotentialof

Canadianskilledimmigrationtoraisepatentrates.

Finally, in the remaining rows of Table 1we report theweighted samplemeans of city-level

averageage,nominalincome,andemploymentrates,aswellastheexpectedpatentspercapitavariable

describedabove.SimplecorrelationswiththesamplemeansinTable1appeartosuggestthatpatenting

ratestendtobehigherinolderpopulationsandtendtoincreaseinrecessions(based,inparticular,on

the large increase in the patenting rate between 1991 and 1996when employment rates fell).More

compelling evidence of these effects is, however, provided by regression analyses that control for

unobservedperiodeffects.

Theresultsfromestimatingequation(1)usingboththeHGLspecification(1)andarichersetof

controls(2)arereportedinTable3.ThefirstcolumnindicatesthatincreasingtheCanadianuniversity-

educatedimmigrantshareby1percentagepointisexpectedtoincreasepatentspercapitabyabout1.1

logpoints.ThecomparableU.S.estimate(seespecification(1)ofTable5inHGL)is14.7logpoints,which

fallsfaroutsidetheconfidenceintervalofourestimate.Thecoefficientonthenativeshareis,however,

almost identical to the HGL estimate (4.5 compared to the HGL estimate of 4.1) and is statistically

significantatthe10%level.Thissuggeststhatthelargedifferenceinourimmigrantshareestimatesdoes

notreflectgreatermeasurementerrorinourpopulationshares,structuraleconomicdifferencesbetween

thetwocountries,orotherdifferencesinourmethodologicaldifferences,suchasourfocusoncities,as

opposedtostates.Infact,ifweuseanalternativespecificationandvariabledefinitionsthatmostclosely

matchthatofHGL,thatis,using10-yearfirst-differences(insteadof5)andcountingpatentsonlyforthe

oneyearfollowingthecensusyearbasedontheresidenceofonlythefirstinventor,thedifferenceinthe

impactofuniversity-educatedimmigrantsacrossthetwocountriesbecomesevenlarger.Althoughthe

variancesoftheestimatedcoefficientsincreasesubstantially,presumablyduetothesmallersamplesize

andnoisierdependentvariable, thepointestimates suggesteven smallerbeneficial impactsof skilled

immigrationinCanada,andaslightlylargerimpactofskillednatives.14

14Theseresultsareavailablefromtheauthorsuponrequest.

14

ThesecondcolumnofTable3presentsourresultsusingarichersetofcontrols.Althoughthe

university-educatedimmigrantcoefficientincreasesto3.5,onparwiththeeffectofuniversity-educated

natives, this coefficient is still statistically insignificant and much smaller than the HGL benchmark

estimate. In the next two columns of Table 3we instead define the skilled population as university-

educatedindividualswhoareemployedinaSTEMoccupation.Asexpected,thepointestimatesincrease

substantially,andthecoefficientsonimmigrantsandnativesarenowsimilarandmuchlarger.Usingthe

HGLcontrols,theestimatedeffectsofincreasingtheskilledimmigrantpopulationsharearenow7.3and

6.3 for immigrants and natives, respectively, but neither estimate is statistically significant. However,

using the richer set of controls increases these estimates to 21.7 and 19.0 and both coefficients are

statisticallysignificantatthe10%level.Takenasawhole,theresultsinTable3appeartosuggestthatthe

impactofuniversity-educated immigrationonCanadianpatentinghasbeenmodestandthat this is in

largepartduetothelowemploymentratesofSTEM-educatedCanadianimmigrantsinSTEMjobs.

In Table 4, we explore this issue in more detail by redefining the skilled population using

informationonfieldofstudy.Sinceweareforcedtodropthe1986-1981differences,were-estimatethe

firsttwocolumnsofTable3usingthesmallersample(columns1and2).Thekeyresultisthatrefiningour

definition of skilled to mean university educated in a STEM field has essentially no impact on the

immigrant coefficient, but increases the native coefficient substantially. Both immigrant coefficients

remainclosetozeroandareinsignificant,whereasthenativecoefficients increaseto16.8and19.1in

specifications (1) and (2), respectively (compared to 5.4 and 4.2 in columns 1 and 2) and are both

significant.ThedifferenceintheimpactofSTEM-educatedimmigrantsandnativesisstark.Anobvious

questionistowhatextentthedifferencereflectstheforeigneducationalcredentialsofimmigrants.Inthe

fifthandsixthcolumnsofTable4,wedistinguishbetweenCanadian-andforeign-educatedimmigrants.

AlthoughtheestimatesforCanadian-educatedimmigrantsarelarger,theyarestillmuchsmallerthanthe

comparablecoefficients fornatives, suggesting that thedifference reflects,at least inpart, something

other than schooling quality. One possible explanation is employer discrimination against Canadian-

educatedimmigrantswithethnicnames,consistentwiththeCanadianauditstudyofOreopoulos(2011).

Finally, inthelasttwocolumnsofTable4weexaminetheimpactofincreasingthepopulation

shareofimmigrantsandnativesthatarenotonlyuniversity-educatedinaSTEMfield,butalsoemployed

inaSTEMoccupation.Hereweseeasubstantialincreaseinthecoefficientontheimmigrantpopulation

share to 9.3 and 36.3 in specifications (1) and (2), respectively. The latter coefficient is statistically

significantat the10% levelandcomparable inmagnitudetothe52.4 for the immigrantscientistsand

15

engineersshare inHGL (Table6panelC).Takenasawhole, theestimatesappear tosuggest that the

relativelysmallcontributionofskilledimmigrantstoinnovationinCanadadoesnotreflecttheeducational

backgroundsofCanadianimmigrants,intermsofeithertheirrelativeconcentrationinSTEMfieldsorthe

qualityof their schooling.Rather, it seems thatbarriers toemployment inSTEM jobsare theprimary

sourceoftheirmodestcontributiontoinnovation.

It is, of course, possible that our naïve FD estimates are downward biased, perhaps as a

consequence of measurement error in the Canadian population shares. In Table 5, we examine the

robustnessofourestimatestoinstrumentingimmigrationtoCanadiancities.AsdescribedinSection3,

weinstrumentchangesinskilledimmigrantpopulationsusingstockpopulationsbasedonCensusdata.

Ourfirststageestimatesaresignificantatthe1%level.

Using our complete sample, we define skilled workers as: (i) the university educated; or (ii)

university-educatedandemployedinaSTEMjob.TheIVestimatesoftheeffectofraisingtheuniversity-

educated immigrant share change little and continue to suggest small positive and statistically

insignificanteffects. This is in sharp contrast toHGL,whoseestimatesbasedon the same instrument

nearlydoubleinmagnitude(seePanelAofTable8).Isolatingtheeffectofincreasingthepopulationshare

comprisedof university-educated immigrantswho are employed in a STEM job continues to produce

substantiallylargerestimates.Usingtherichercontrols(specification2)thepointestimategoesfrom1.1

to10.4andisstatisticallyinsignificant,althoughthelatterisnowhalfwhatitwasinTable3..15

5 Conclusions

WearguethatCanadaisanimportantcasestudybecauseits`pointssystem’forscreeningprospective

immigrantsisseenbymanyasamodelofhowtoraisetheaverageskilllevelsofimmigrationinflows.The

main finding from our analysis is that Canadian STEM-educated immigrants who are successful in

obtainingjobsinSTEMareasdoappeartoraisepatentingratesinasignificantway.However,withlittle

more than one-third of STEM-educated immigrants finding employment in STEM jobs, the impact of

Canadian skilled immigrationonpatent rateshasbeen relativelymodest in comparison to theUnited

States.ThefactthattheemploymentratesofCanadianSTEM-educatedimmigrantsinSTEMjobshas,if

anything, tended to decrease over time, while the comparable rate for Canadian natives has been

15Afurtherconcernisthattheinclusionofendogenouscontrolvariablescouldbiasourresults.WerantheIV

specificationsintable5withonlyfixedeffectsandobtainedsimilarcoefficientsfortheshareofuniversity-

educatedimmigrantsandsomewhatlargerbutstillinsignificantcoefficientsonuniversity-educatedstem-

employedimmigrantshares.

16

increasing,shouldbecauseforconcernamongpolicymakerscontemplatingintroducing‘pointssystems’

forimmigrantselection.Giventhemodestmagnitudeofourestimatedeffects,itappearsthat,forCanada,

anyspillovereffectsofimmigrantsonnativepatentingareminimal.

Whatisthepolicyrelevanceofthesefindings?Itwouldappearthatadoptinga`pointssystem’so

astoputmoreweightonSTEMeducationalbackgroundsisunlikelytohavethedesiredeffectofboosting

innovation.Rather,ourevidenceemphasizesthatselectingimmigrantswithSTEMskillsisnotsufficient,

giventhechallengesthatCanadianSTEM-educated immigrantsappeartoface inobtainingSTEMjobs.

The critical question for policy is whether the employment barriers that STEM-educated immigrants

appeartofacereflectdifferencesintheirskillsandabilitiesorlabourmarketinefficienciesarisingfrom

informationfrictionsinjobsearch,foreigncredentialassessment,orracialdiscrimination.Inthisregard,

itisnoteworthythatSTEM-educatedimmigrantsfindSTEMemploymentlessfrequentlythannativeseven

when they were educated in Canadian universities and that the contribution of STEM-educated

immigrantsfromCanadianuniversitiesappearstoalsofall farshortofthecomparablecontributionto

innovationofnative-bornCanadians.Thissuggeststousthatmorethaninformationfrictionsaroundthe

valueofimmigrants’educationalcredentialsisresponsible.

An alternative explanation is that the employment challenges of Canadian STEM-educated

immigrants primarily reflect differences in Canadian andU.S. skilled immigration policy. In particular,

whereasthevastmajorityofskilledimmigrantsintheU.S.areadmittedviatemporaryworkpermitsfrom

sponsoring employers, H-1B visas in particular, skilled-stream immigrants arriving in Canada as new

permanent residents typically do not have pre-arranged employment.16 Instead, theCanadian `points

system’hashistoricallygrantedpermanent residency to foreignapplicants solelyon thebasisof their

foreigneducational credentialsandyearsofworkexperience.To theextent thatU.S.employershave

richerinformationregardingtheproductivityofforeignworkers,STEM-educatedimmigrantsintheU.S.

arenotonlymorelikelyto“hitthegroundrunning”withajob,butmayalsobeofhigherlabourmarket

“quality”ondimensionsunobservabletothe`pointssystem.’Thissuggeststhatagreateremphasison

pre-arrangedemploymentinimmigrantselectioncouldbebeneficial.Indeed,thepastdecadehasseen

the introductionofanumberofnewskilled immigrationprogramseasingthetransitiontopermanent

16AdministrativedatafromtheU.S.OfficeofImmigrationStatisticsindicatethatsomewherebetween75%and

90%ofnewskill-streampermanentresidentsintheU.S.between2001and2011transitionedfromatemporary

workpermitorstudentvisa(seeYearbookofImmigrationStatistics,HomelandSecurity,variousyears).Incontrast,

overthesameperiod,between10%and25%ofCanadianskilled-streamimmigrantstransitionedfromaworkor

studentvisa(seeFactsandFigures,Immigration,Refugees,andCitizenshipCanada,variousyears).

17

residencyforindividualswithCanadianworkexperienceandjoboffersfromCanadianemployers.17Time

willtellwhethertheseprogramshavebeensuccessfulinraisingtheSTEM-employmentratesofCanada’s

STEM-educatedimmigrantsand,inturn,thecontributionofCanadianimmigrationtoinnovation.

17TheseprogramsincludetheCanadianExperienceClassprogramintroducedin2008,ProvincialNominee

Programs(Ontariowasthelastprovincetointroduceaprogramin2007),andtheExpressEntrySystemfor

processingapplicationsforpermanentresidency,whichwasintroducedin2015.

18

References

Beckstead,D,andG.Gellatly (2006), “Innovationcapabilities: scienceandengineeringemployment in

CanadaandtheUnitedStates,”StatisticsCanadaResearchPaper.

Card, D. (2001), “Immigrant Inflows, Native Outflows, and the Local Market Impacts of Higher

Immigration,”JournalofLaborEconomics19(1):22-64.

Carley,M.,D.Hegde,andA.Marco(2013),“WhatistheprobabilityofreceivingaU.S.patent?,”USPTO

EconomicWorkingPaper.

Chellaraj,G.,K.E.Maskus,andA.Mattoo(2008),“TheContributionofInternationalGraduateStudents

toUSInnovation,”ReviewofInternationalEconomics16(3):444–62.

Clarke,A.andM.Skuterud(2013),“WhyDoImmigrantWorkersinAustraliaPerformBetterThanThose

inCanada?IsittheImmigrantsorTheirLabourMarkets?”CanadianJournalofEconomics46(4):1431-1462.

Clarke,A.andM.Skuterud(2016),“AComparativeAnalysisof ImmigrantSkillsandTheirUtilization in

Australia,Canada,andtheUSA”JournalofPopulationEconomics29(3):849-882.

Ely,R.andD.Thomas(2001),“CulturalDiversityatWork:TheEffectsofDiversityPerspectivesonWork

GroupsProcessesandOutcomes,”AdministrativeScienceQuarterly46(2):229-273.

Green A. and D. Green (2004), “The Goals of Canada’s Immigration Policy: A Historical Perspective,”

CanadianJournalofUrbanResearch13(1):102-139.

Green,DavidandChrisWorswick(2012),“Immigrantearningsprofilesinthepresenceofhumancapital

investment:Measuringcohortandmacroeffects,”LabourEconomics19:241-259.

Herring,C. (2009), “DoesDiversityPay?Race,Gender,and theBusinessCase forDiversity,”AmericanSociologicalReview74(2):208-224.

Hunt, J. andM.Gauthier-Loiselle (2010), “HowMuchDoes ImmigrationBoost Innovation?”AmericanEconomicJournal:Macroeconomics2:31-56.

KerrW. andW. Lincoln (2010). “The Supply Side of Innovation: H – 1B Visa Reforms andU.S. Ethnic

Invention,”JournalofLaborEconomics28(3):473-508.

NathanM. (2014a),“SameDifference?MinorityEthnic Inventors,Diversityand Innovation in theUK,”

JournalofEconomicGeography15(1):129-168.

Nathan,M.(2014b),“TheWiderEconomicImpactsofHigh-SkilledMigrants:ASurveyoftheLiteraturefor

ReceivingCountries,”IZAJournalofMigration3(4):1-20.

Oreopoulos,P.(2011),“WhyDoSkilledImmigrantsStruggleintheLaborMarket?AFieldExperimentwith

ThirteenThousandResumes,”AmericanEconomicJournal:EconomicPolicy3:148-171.

Ozgen C., P. Nijkamp, and J. Poot, (2012), “Immigration and Innovation in European Regions,” in P.

Nijkamp, J. Poot, M. Sahin (eds.), Migration Impact Assessment: New Horizons, Edward Elgar,Cheltenham.

19

Parrotta, P., D. Pozzoli, and M. Pytlikova (2014), “The Nexus Between Labor Diversity and Firm’s

Innovation,”JournalofPopulationEconomics27:303-364.

Peri, G. (2007), “Higher Education, Innovation and Growth,” in Education and Training in Europe, G.Brunello,P.Garibaldi,andEtienneWasmer(eds.),Oxford:OxfordUniversityPress,56-70.

Picot,G.andA.Sweetman(2012),“MakingItinCanada:ImmigrationOutcomesandPolicies,”IRPPStudy,no.29.

Richard,O.,A.McMillan,K.Chadwick,andS.Dwyer(2003),“EmployinganInnovationStrategyinRacially

DiverseWorkforces,”GroupandOrganizationManagement28(1):107-126.

Skuterud,M.andM.Su(2012),““TheInfluenceofMeasurementErrorandUnobservedHeterogeneityin

Estimating Immigrant Returns to Foreign and Host-Country Sources of Human Capital,” EmpiricalEconomics43(3):1109-1141.

VanKnippenberg,D.,DeDreu,C.,andA.Homan(2004),“WorkGroupDiversityandGroupPerformance:

AnIntegrativeModelandResearchAgenda,”JournalofAppliedPsychology89(6):1008-1022.

Xu, Li (2012), “Who Drives a Taxi in Canada?” Research and Evaluation Branch, Citizenship and

ImmigrationCanada.

Figure1:University-educatedimmigrantpopulationsharesandpatentspercapita,CanadaandtheUSA,1980-2006

Notes:ForeachofCanadaandtheU.S.thefigurepresentstheshareofthepopulationaged25andoverthatiscomprisedofuniversity-educatedimmigrants

(lefthandsideaxis)andthenumberofUSPTOpatentsgrantedtoCanadianandU.S.inventorsper100,000population(righthandsideaxis).Forthelatter

series,theyearistheapplicationyearofthepatent.Fractionalpatentswereawardedtoeachcountrywhenthepatenthadmultipleinventorsfromdifferent

countries.OnlypatentsgranteduptoNovember2014weretabulated.Forbothcountries,boththeshareofUniversity-educatedimmigrantsandpatentsper

capitashowanoverallincrease.

010

2030

4050

Pate

nts

per c

apita

(x 1

00,0

00)

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7U

nive

rsity

-edu

cate

d im

mm

igra

nt s

hare

1980 1986 1991 1996 2001 2006

CA immigrants US immigrants CA patents US patents

21

Table1:Population-weightedsamplemeansbyCensusyear

1981 1986 1991 1996 2001 2006 2006–1981/6

difference

Patents 489.7 744.2 1055.2 1553.1 1755.9 1668.7 1179.0***

Patentspercapita(x100,000) 42.2 58.5 74.0 105.8 113.2 103.0 60.8***

Population 971,384 1,074,428 1,169,049 1,277,834 1,383,794 1,504,691 533,307***

Immigrantpopulationshare 0.223 0.219 0.231 0.247 0.255 0.268 0.046**

- Universityeducated 0.027 0.030 0.037 0.047 0.060 0.076 0.050***

- UniversitySTEMeducated -- 0.010 0.012 0.016 0.022 0.030 0.020***

- Canadian-universitySTEMeducated -- 0.005 0.006 0.008 0.10 0.013 0.008***

- Foreign-universitySTEMeducated -- 0.005 0.006 0.008 0.012 0.017 0.012***

- Universityeducated&STEMemployed 0.004 0.004 0.005 0.006 0.009 0.011 0.007***

- UniversitySTEMeducated&STEMemployed 0.003 0.004 0.005 0.008 0.010 0.006***

Native-bornpopulationshare 0.777 0.781 0.769 0.753 0.745 0.732 -0.046**

- Universityeducated 0.073 0.087 0.102 0.115 0.128 0.142 0.069***

- UniversitySTEMeducated -- 0.019 0.021 0.022 0.025 0.027 0.008***

- Universityeducated&STEMemployed 0.007 0.008 0.009 0.010 0.013 0.014 0.007***

- UniversitySTEMeducated&STEMemployed 0.006 0.007 0.008 0.009 0.010 0.004***

Meanage 32.6 33.7 34.6 35.4 36.7 38.0 5.3***

Meanincome 9222 13,398 18,385 19,430 24,032 28,947 19,725***

Employmentrate 0.659 0.657 0.672 0.652 0.688 0.700 0.041***

Expectedpatentspercapita(x100,000) 42.2 58.4 73.9 105.7 113.1 102.9 60.7***

Observations 98 98 98 98 98 98 196

Notes:SamplemeansofvariablesusedintheregressionsbyCensusyear.Themeansareweightedbycitypopulationssothattheyarerepresentativeofthe

CanadianpopulationresidingwithinoneofCanada’slargest98cities.PatentsarethecumulativesumofannualpatentsinthefiveyearsfollowingeachCensus

year.Populationsharesarecalculatedasthefractionofindividualsaged15-70thatfallineachcategory.The1981CanadianCensusdoesnotreportfieldof

study.Incomesarenotdeflated.Expectedpatentspercapitacontrolsforthenumberofpatentsthateachcitywouldhavebasedonitsdistributionof1972-

1980patentsacrossdifferentpatentclassesandthesubsequentnational-levelgrowthinthenumberofpatentsforeachofthosepatentclassesacrossCensus

years.*p<.10,**p<.05,***p<.01.

22

Table2:ConditionalprobabilitiesofSTEMeducationandSTEMemploymentforimmigrantsandnatives

1986 1991 1996 2001 2006 2006–1981

difference

Immigrants

Pr[Universityeducated] 0.138 0.160 0.188 0.233 0.285 0.165

Pr[STEMeducated|universityeducated] 0.324 0.328 0.337 0.374 0.387 0.062

Pr[Canadianeducation|STEMuniversityeducated] 0.505 0.519 0.500 0.447 0.429 -0.076

Pr[Foreigneducation|STEMuniversityeducated] 0.495 0.481 0.500 0.553 0.571 0.076

Pr[STEMemployed|STEMuniversityeducated] 0.348 0.338 0.311 0.345 0.322 -0.026

Pr[STEMemployed|CanadianSTEMuniversityeducated] 0.363 0.355 0.322 0.343 0.343 -0.019

Pr[STEMemployed|ForeignSTEMuniversityeducated] 0.333 0.320 0.301 0.347 0.307 -0.027

Natives

Pr[Universityeducated] 0.112 0.132 0.153 0.172 0.194 0.101

Pr[STEMeducated|universityeducated] 0.214 0.202 0.193 0.195 0.191 -0.023

Pr[STEMemployed|STEMuniversityeducated] 0.342 0.355 0.355 0.370 0.370 0.028

Notes:Conditionalprobabilitiesconstructedusingthemeanpopulationshares(weightedbypopulationsize)forindividualsaged15-70acrossCanada’slargest

98cities.Canadianimmigrantsaremorelikelytoholdauniversitydegreethantheirnative-borncounterpartsandthisdifferencehasgrownovertime.

University-educatedimmigrantsarealsomorelikelytobeeducatedinaSTEMfieldthantheirnative-borncounterpartsandthisdifferencehasalsobecome

largerovertime.However,theprobabilityofaSTEM-university-educatedimmigrantbeingemployedinaSTEMoccupationhastendedtodecreaseovertime,

whereasithasincreasedfornatives.Thissuggestsaneducation-jobmismatchforimmigrants.

23

Table3:WLS-FDestimatesoftheeffectofuniversity-educatedanduniversity-educated-STEM-employedimmigrantpopulationsharesonlogpatentspercapita

University-educated University-educated&STEM-employed

(1) (2) (1) (2)

Immigrantpopulationshare 1.118

(1.677)

3.508

(2.992)

7.276

(7.915)

21.743*

(12.887)

Nativepopulationshare

4.457*

(2.391)

3.315

(3.303)

6.328

(9.318)

19.007*

(11.415)

Logmeanage 0.494

(1.257)

-0.589

(1.507)

0.772

(1.228)

-0.431

(1.452)

Logpopulation(1981) 0.003

(0.008)

-0.006

(0.012)

0.006

(0.008)

-0.009

(0.013)

Logmeanincome(1981) 0.053

(0.112)

-- 0.005

(0.113)

--

Logmeanincome -- -0.028

(0.607)

-- -0.521

(0.630)

Employmentrate -- -0.094

(1.266)

-- -0.034

(1.275)

Logexpectedpatentspercapita -- 0.202*

(0.116)

-- 0.231*

(0.118)

Yearfixedeffects Yes No Yes No

Year-regionfixedeffects No Yes No Yes

R-squared 0.285 0.332 0.284 0.340

Numberofobservations 490 490 490 490

Notes:Weightedleastsquares(observationsareweightedbycitypopulation)regressionswithfiveyearfirstdifferences.Thedependentvariableisthelogof

patentspercapita.Standarderrorsareclusteredbycity.ThesampleconsistsofCanada’s98largestcitiesand1986-1981,1991-1986,1996-1991,2001-1996,

and2006-2001firstdifferences.Specification(1)attemptstorecreatethespecificationinHGLwhilespecification(2)includesarichersetofcontrols.*p<.10,

**p<.05,***p<.01.

24

Table4:WLS-FDestimatesoftheeffectofuniversity-educated,university-STEM-educated,anduniversity-education-STEM-employedimmigrantpopulation

sharesonlogpatentspercapita

University-educated University-STEM-educated University-STEM-educated University-STEM-educated

&STEM-employed

(1) (2) (1) (2) (1) (2) (1) (2)

Immigrantpopulationshare -1.026

(1.800)

0.511

(3.417)

-3.342

(3.628)

1.093

(4.855)

-- -- 9.265

(13.658)

36.341*

(19.855)

ImmigrantCanadianuniversity -- -- -- -- 4.295

(29.814)

3.406

(42.164)

-- --

Immigrantforeignuniversity -- -- -- -- -5.686

(8.282)

0.309

(13.952)

-- --

Nativepopulationshare 5.389*

(3.096)

4.156

(4.210)

16.784*

(9.148)

19.109*

(10.661)

16.525*

(9.112)

19.013*

(10.340)

17.563

(16.666)

26.522

(20.611)

Logmeanage -.260

(1.476)

-1.814

(1.801)

-0.452

(1.397)

-1.825

(1.709)

-0.357

(1.349)

-1.817

(1.676)

0.456

(1.428)

-1.331

(1.714)

Logpopulation(1981) 0.020**

(0.010)

0.009

(0.014)

0.020*

(0.11)

0.007

(0.013)

0.018

(0.015)

0.007

(0.019)

0.016

(0.010)

-0.013

(0.016)

Logmeanincome(1981) 0.072

(0.119)

-- 0.072

(0.120)

-- 0.066

(0.122)

-- -0.034

(0.126)

--

Logmeanincome

-- -0.166

(0.649)

-- -0.258

0.608

-- -0.261

(0.597)

-- -0.874

(0.697)

Employmentrate --

-0.929

(1.314)

-- -1.045

(1.307)

-- -1.027

(1.251)

-- -0.632

(1.337)

Logexpectedpatentspercapita --

0.147

(0.117)

-- 0.154

(0.119)

-- 0.153

(0.116)

-- 0.181

(0.123)

Yearfixedeffects Yes No Yes No Yes No Yes No

Year-regionfixedeffects No Yes No Yes No Yes No Yes

R-squared 0.253 0.284 0.253 0.287 0.254 0.287 0.250 0.297

Numberofobservations

392 392 392 392 392 392 392 392

Notes:Weightedleastsquares(observationsareweightedbycitypopulation)regressionswithfive-yearfirstdifferences.Thedependentvariableisthelogof

patentspercapita.Standarderrorsareclusteredbycity.ThesampleconsistsofCanada’s98largestcitiesand1991-1986,1996-1991,2001-1996,and2006-

2001first-differences,sincefieldofstudyinformationisnotavailableinthe1981Census.Specifications(1)attempttorecreatethespecificationinHGLwhile

specifications(2)includearichersetofcontrols.*p<.10,**p<.05,***p<.01

25

Table5:IV(2SLS)estimatesoftheeffectofuniversity-educatedanduniversity-educated-STEM-employedimmigrantpopulationsharesonlogpatentsper

capita

University-educated University-educated&STEM-employed

(1) (2) (1) (2)

Immigrantpopulationshare 2.870

(4.393)

1.060

(5.656)

8.730

(11.347)

10.404

(13.912)

Nativepopulationshare

5.350

(3.650)

2.006

(4.288)

5.964

(8.981)

20.388*

(11.115)

Logmeanage 0.733

(1.409)

-0.725

(1.451)

0.828

(1.294)

-0.549

(1.400)

Logpopulation(1981) -0.002

(0.015)

0.004

(0.022)

0.006

(0.009)

-0.001

(0.011)

Logmeanincome(1981) 0.027

(0.107)

-- -0.001

(0.113)

--

Logmeanincome -- 0.121

(0.613)

-- -0.259

(0.580)

Employmentrate --

-0.235

(1.251)

-- -0.212

(1.257)

Logexpectedpatentspercapita --

0.202*

(0.111)

-- 0.221**

(0.109)

Yearfixedeffects Yes No Yes No

Year-regionfixedeffects No Yes No Yes

R-squared 0.285 0.331 0.284 0.338

Numberofobservations 490 490 490 490

Firststage:

Exp.ImmigrantShare

0.622***

(0.157)

0.588***

(0.144)

0.943***

(0.327)

0.908***

(0.194)

Fstatistic 64.90 345.39 32.01 189.69

Notes:Weinstrumentlocalskilledimmigrantpopulationsusingpredictedimmigrantpopulationsbasedonthehistoricalcity-levelsettlementpatternsof

immigrantsfromparticularorigincountriesandnational-levelpopulationsofimmigrantsfromthosecountries.Estimatesarefromtwo-stageleastsquare.

Observationsareweightedbycitypopulation.Standarderrorsareclusteredbycity.ThesampleconsistsofCanada’s98largestcitiesand1986-1981,1991-

1986,1996-1991,2001-1996,and2006-2001firstdifferences.Specifications(1)attempttorecreatethespecificationinHGLwhilespecifications(2)includea

richersetofcontrols.*p<.10,**p<.05,***p<.01.

26

Appendix

STEMfieldsofstudyintheCanadianCensusdataareidentifiedusinginformationonmajorfieldofstudy(MFS),whichisidentifiedforall

individualswhohavecompletedapost-secondaryprogramofstudy.MajorfieldofstudyiscodedusingaMFSclassificationsystemduringthe

censusyears1986,1991,1996and2001,whilein2006itiscodedaccordingClassificationofInstructionalProgram(CIP)Canada2000.

Therefore,weusetheMFSclassificationasthemastercodeandmaptheCIPtoMFS,andthenselectthestudyfieldsfromMFStoidentifySTEM

fields.

ToconstructaconcordancebetweenMFSandCIP,wemakeuseoftheempiricalconcordancesfromCIPtoMFSprovidedbyStatisticsofCanada

(http://www12.statcan.ca/census-recensement/2006/ref/dict/app-ann020-eng.cfm).Theempiricalconcordancesprovidemappingsofthe

distributionalrelationshipsbetweenthetwoclassifications.Thedetailsaredescribedonthewebsite.TherearethreelevelsofMFSandCIP

groupingsrespectively,correspondingly,threeconcordancesareprovidedforeachgrouplevel:CIPprimarygroupings-MFSmajorlevel(level1),

CIPsubseries(4digit)andMFSminorlevel(level2),andCIPinstructionalprograms(6digit)andMFSunitlevel(level3).Intheseconcordances,

asharevariableiscalculatedasthepercentageoftotalCIPthatisaccountedforbythespecificMFScode.ThusforeachCIP,thesharesaddup

to1.AhighersharevalueindicatesamorefrequentoccurringofaMFSinaCIP.

OurstrategyistotakethesharevariableforeachCIPandapplythemodemethod.Inparticular,westartfromthelevel3concordance(theleast

aggregatedcategories),andmapaCIPtoaMFSwhichreturnsahighestsharevaluegiventhatparticularCIP.IftherearesomeCIPcategories

notmappedtoMFSinlevel3concordances,wethenusethelevel2concordancesandapplythesamemethod,andthenlevel1(Atlast,there

arequitefewCIPcategoriesnotbeingmapped,wethenreadthedescriptionsonthoseCIPvariablesandmapthemtoMFSmanually.).Alistof

theconcordanceisprovidedinTable3.Consequently,theSTEMfieldismadeupbyfourmajorMFScategories:‘Agricultural,biological,

nutritionalandfoodsciences’,‘Engineeringandappliedsciences’,‘Appliedsciencestechnologiesandtrade;Mathematics’,and‘Computerand

physicalsciences’.

TheSTEMoccupationvariableisconstructedbasedontheoccupationinformationineachcensusfile.Tobespecific,1980standardoccupational

classification(occ81)systemisusedin1981and1986censusfilesrespectively,and1991standardoccupationalclassification(soc91)systemis

usedin1991,1996,2001and2006censusfilesrespectively.Accordingly,in1981and1986censusfiles,theSTEMoccupationisidentifiedifthe

variableocc81fallsintothecategory‘MajorGroup21–OccupationsinNaturalsciences,engineeringandmathematics’;whileintherestcensus

files,theSTEMoccupationisidentifiedifthevariablesoc91fallsintothecategory‘C-NaturalandAppliedSciencesandRelatedOccupations’.

27

TableA1:SamplewithoutToronto-FDestimatesoftheeffectofuniversity-educatedanduniversity-educated-STEM-employedimmigrantpopulationshares

onlogpatentspercapita

University-educated University-educated&STEM-employed

(1) (2) (1) (2)

Immigrantpopulationshare 2.838

(2.956)

7.733*

(4.518)

26.339*

(15.077)

42.861***

(15.573)

Nativepopulationshare

3.831

(2.789)

2.917

(3.281)

-5.397

(12.961)

10.906

(13.532)

Logmeanage 0.305

(1.222)

-0.212

(1.448)

0.665

(1.157)

-0.082

(1.391)

Logpopulation(1981) 0.008

(0.010)

-0.008

(0.014)

0.015

(0.010)

-0.006

(0.011)

Logmeanincome(1981) 0.031

(0.122)

-- -0.046

(0.121)

--

Logmeanincome -- -0.490

(0.688)

-- -1.130

(0.764)

Employmentrate -- 0.746

(1.313)

-- 1.094

(1.253)

Logexpectedpatentspercapita -- 0.167

(0.109)

-- 0.196*

(0.109)

Yearfixedeffects Yes No Yes No

Year-regionfixedeffects No Yes No Yes

R-squared 0.259 0.305 0.263 0.318

Numberofobservations 485 485 485 485

Notes:Weightedleastsquares(observationsareweightedbycitypopulation)regressionswithfiveyearfirstdifferences.Thedependentvariableisthelogof

patentspercapita.Standarderrorsareclusteredbycity.ThesampleconsistsofCanada’s97largestcitieswiththeexceptionofTorontoand1986-1981,1991-

1986,1996-1991,2001-1996,and2006-2001firstdifferences.Specifications(1)attempttorecreatethespecificationinHGLwhilespecifications(2)includea

richersetofcontrols.

*p<.10,**p<.05,***p<.01

28

TableA2:Sampleofcitieswithapopulationofatleast40,000in1981–UnweightedFDestimatesoftheeffectofuniversity-educatedanduniversity-educated-

STEM-employedimmigrantpopulationsharesonlogpatentspercapita

University-educated University-educated&STEM-employed

(1) (2) (1) (2)

Immigrantpopulationshare 7.146

(6.062)

8.788

(6.527)

38.742*

(19.393)

30.638

(20.571)

Nativepopulationshare

8.605**

(4.256)

5.191

(5.473)

20.399

(23.722)

29.476

(26.300)

Logmeanage -0.330

(1.898)

-1.782

(2.035)

0.113

(1.931)

-1.495

(2.119)

Logpopulation(1981) -0.004

(0.022)

-0.024

(0.023)

0.007

(0.017)

-0.013

(0.014)

Logmeanincome(1981) -0.260

(0.244)

-- -0.299

(0.237)

--

Logmeanincome -- 1.004

(0.825)

-- 0.760

(0.880)

Employmentrate -- -1.027

(2.300)

-- -0.995

(2.243)

Logexpectedpatentspercapita -- 0.116

(0.140)

-- 0.146

(0.144)

Yearfixedeffects Yes No Yes No

Year-regionfixedeffects No Yes No Yes

R-squared 0.241 0.346 0.243 0.352

Numberofobservations 265 265 265 265

Notes:OLSfirst-differenceregressions.Thedependentvariableisthelogofpatentspercapita.Standarderrorsareclusteredbycity.Thesampleconsistsof

the53citieswithapopulationofatleast40,000in1981and1986-1981,1991-1986,1996-1991,2001-1996,and2006-2001firstdifferences.Specifications(1)

attempttorecreatethespecificationinHGLwhilespecifications(2)includearichersetofcontrols.

*p<.10,**p<.05,***p<.01