Está en la página 1de 18

Canadian Public Policy

Access to College and University: Does Distance to School Matter?


Author(s): Marc Frenette
Source: Canadian Public Policy / Analyse de Politiques, Vol. 30, No. 4 (Dec., 2004), pp. 427-443
Published by: University of Toronto Press on behalf of Canadian Public Policy
Stable URL: http://www.jstor.org/stable/3552523 .
Accessed: 31/03/2011 18:21

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=utp. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

University of Toronto Press and Canadian Public Policy are collaborating with JSTOR to digitize, preserve
and extend access to Canadian Public Policy / Analyse de Politiques.

http://www.jstor.org
Access to College and University:
Does Distance to School Matter?
MARC FRENETTE
Businessand LabourMarketAnalysisDivision
StatisticsCanada
Ottawa,Ontario

On sait ddj~quepoursuivredes dtudesau-deladu secondaireest moinsprobablechez les dtudiantsissus de


famillesafaible revenu.Nous apportonsune contributiona l'dtudede l'accbsaux dtudespost-secondaires
en examinantle r61equejoue la distancequi sdpareles dtudiantsde l'dcole secondairedes dtablissements
d'enseignementpost-secondaire,et en pretantune attentionparticulibreau choix que font les dtudiants
entrele collbgeet l'universitd.La distancede l'dcole peutdissuaderl'dtudiantde poursuivreses dtudes,en
raisonducofitdurelogement,surtoutdansle cas de l'dtudiantissu d'unefamilleafaiblerevenu.Les rdsultats
de notredtudesoutiennentl'idde que, al'accroissementde la distancede l'dcole s'associe un decart dans
1'accbsaux 6tudespost-secondaires,dcartqui va bien au-delade celui qui a 6tdlid au revenufamilial.Plus
prdcisdment,l'accroissementde la distancede l'universitdest associdea une frdquentation plus faible de
a
l'universitdet une plus grandetendance frdquenter le collbgelocal. Ce phdnombnes'avbrebeaucoupplus
marquddansles famillesa faible revenu.
It is alreadyknownthatstudentsfromlower-incomefamiliesareless likely to pursuepostsecondarystudies.
This study contributesto our knowledgeof postsecondaryaccess by focusing on the role played by the
distanceseparatinghigh-schoolstudentsfrompostsecondaryinstitutions,withparticularfocuson thechoice
betweencollege anduniversityattendance.Distanceto school may act as a deterrentto attendingby virtue
of relocationcosts, especiallyif the studentis froma lower-incomefamily.The studyfindingssupportthe
notionthatincreaseddistanceto school is associatedwith an access gap, even aboveandbeyondthe gap
thathas beenlinkedto familyincome.Specifically,increaseddistanceto universityis associatedwith lower
universityattendance,anda largertendencyto attendthe local college instead.This relationshipis foundto
be muchstrongerin lower-incomefamilies.

INTRODUCTION important implications for earnings inequality,


intergenerationalincome mobility,and skill devel-
economicvalueof a or
university college opment, some more recent work has focused on
The
educationhas long been studiedby economists issues concerningaccess to postsecondaryschool-
and sociologists.The literaturehas reacheda con- ing. This line of research has very clear policy
sensuson the averagerateof returnto an additional relevancefor studentloan programsand scholar-
year of schooling- roughly6 percentto 10 per- ships,especiallythosethatareat leastpartiallybased
cent in additionalearnings.1Becauseit may have on financialneed.An obviousmeasureof financial

CANADIANPUBLICPOLICY- ANALYSEDE POLITIQUES,


VOL.XXX,NO.4 2004
428 Marc Frenette

need is familyincome,which is knownto be posi- Moreover,the negativeeffect of distanceis felt far


tively correlatedwith access to university (e.g., more amongstudentsfromlower-incomefamilies,
Mehmet1978;MengandSentence1982;Christofides, which supportsthe notionthatfinancialcosts are a
CirelloandHoy 2001; andCorak,LippsandZhao largepartof thereasonwhydistanceis sucha strong
2003). deterrentto attendinguniversity.AndresandLooker
(2001) also reachedsimilarconclusionswith dif-
Some studentsmay also face geographicbarri- ferentdataandmethods.
ers, which may translateto financialneed because
of the addedcosts of movingawayto attendschool. The currentstudyfollows up on the distanceto
These includedirectfinancialcosts (e.g., rentinga schoolworkby Frenette(2005)withmorecomplete
movingvan, rentingan apartment,purchasingfur- datathatincludesinformationon colleges. First,do
nitureor otheritemsfortheapartment, etc.), as well studentswholive too farto attenduniversity"make
as indirectfinancialcosts (e.g., foregoneeconomies up"forthisdisadvantageby attendingcollege?Sec-
of scaleassociatedwithdividingthefamilyin two2). ond, is this uptakein college participationlikely to
Theremay also be emotionalcosts associatedwith occuramongstudentsfromlower-income families-
leaving home. Some studentsmay be reluctantto those most negativelyaffectedby living far away
leave homebecauseof theirnetworkof familyand from universities?Living too far to attendcollege
friends,or they may simplyfeel unprepared to live is generallyless of anissue, giventhatcollegeshave
on theirown. Of course,for otherstudents,added a strongruralpresence.Nevertheless,a thirdobjec-
distancebetweenthe home and the school may be tive of the paperis to assess the role of commuting
seen as an advantageif they desiremoreindepend- distancein college participation.
ence fromtheirparents.
Studyingthe role of distanceto schoolis appeal-
Identifyingthe magnitudeof these geographic ing from an analyticalpoint of view comparedto
barriers,theirrolein shapingpostsecondary partici- other cost-relatedbarriers.For example,studying
pationpatterns,and who is most affectedby these the role of tuitionfees in postsecondaryaccess can
barrierscan provideuseful informationfor the de- be problematicforseveralreasons.First,tuitionfees
velopment of needs-based student financial are not necessarilyexogenoussince they may re-
assistanceprograms.Thesegeographicbarriersmay spondto anincreaseddemandfortheprogram.This
also be linkedto skill shortagesin ruralareas.Post- wouldleadto a positivecorrelationbetweentuition
secondarygraduateswho werebornin a givenrural and programparticipation,whichmay be counter-
areamay be less reluctantto settle thereaftertheir intuitive.Moreover,exogenouschangesin tuition
studies.Despitethis, verylittle is knownaboutdis- fees may be largelydeterminedby policy changes
tance to school andits potentialrole in explaining at the federalor provinciallevel, ratherthanat the
postsecondaryparticipationpatterns. institutionorprogramlevel.Thislargelyreducesthe
numberof naturalor randomexperimentsthat re-
Some recent work by Frenette(2005) suggests searchers could draw upon.3 On the other hand,
that distanceto school plays an importantrole in differencesin distanceto school areperhapscloser
As manyas one in fivehigh-
universityparticipation. to randomexperiments.By andlarge,peopledo not
school studentslive beyond80 km of straight-line choose wherethey growup. Whatmightmakedis-
distancefroma university,andareonly 58 percent tance to school less random are significant
as likelyto attenduniversityas studentslivingwithin differencesin the cognitiveabilitiesof childrenin
easy commutingdistance- less than40 km (after urbanandruralareas.In this study,parentaleduca-
accountingfor differencesin familyincome,paren- tion and family income are used to proxy the
tal educational attainment, sex, and province). cognitive abilities of the parents,which may be

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Accessto Collegeand University:DoesDistanceto SchoolMatter? 429

highly correlatedwith the cognitiveabilitiesof the be addressedthroughstudentfinancialassistance


students. programstargetedat studentsin outlyingareas(i.e.,
student loans, bursariesor grants)or by funding
Briefly, the results indicatethat thereis a con- morespacesin new or existinginstitutionsor satel-
siderable uptake in college participationamong lite campusesin outlyingareas(i.e., buildingnew
studentsliving neara college only (relativeto stu- structures,hiringadditionalinstructors,and subsi-
dents living near a universityand a college). This dizing tuition). In other words, do we bring the
uptakein college participationalmost fully coun- student to the school or bring the school to the
terbalancesthe lower universityparticipationrate student?
amongstudentsliving neara college only,suchthat
the overallpostsecondaryparticipation rate(univer- The fact thatstudentscan be easily classifiedby
sity and college) is about the same for bothgroups their proximityto postsecondaryinstitutionspro-
of students- about40 percenteach.The uptakeis vides a convenient mechanism for increasing
concentratedamongstudentsfromlower-andmid- fundingdirectlythroughstudentassistance.Know-
dle-income families - those most negatively ing themagnitudeof thegeographicaccessgap,and
affectedby distancein accessinguniversity.Students which studentsare most affected by it, provides
from upper-incomefamilies show no signs of an valuable information in the design of student-
uptakein college participationwhenlivingfaraway assistanceprograms.
fromuniversity.Studentsliving beyondcommuting
distancefrom college are 37 percentless likely to Increasingfundingvia new or existing institu-
attendcollege thanthose living withincommuting tionsin outlyingareasis a slightlymorecomplicated
distance.Studentsfromlower-incomefamilies are matter.First, thereis the issue of local demandin
affectedfar moreby living beyondcommutingdis- these areas.Not only are these communitiesquite
tance to college than are other students. It is small,butthe groupof studentsthatwouldpresum-
importantto note, however,thatonly about3 per- ably benefitthe most from additionallocal spaces
centof high-schoolstudentslive beyondcommuting is also very small(17 percentof all studentsdo not
distanceto college. have local access to a university,andof these, it is
studentsfromfamiliesin thebottomincometierthat
The balance of this paper follows in standard aremost affectedby distance).It is also not certain
fashion.The next sectiondescribesthe policy con- thatadditionalspaceswouldbe filled by local stu-
text surroundingissues of geographic access to dents,as excess demandmayexist in largercentres.
postsecondaryinstitutions.This is followed by the In thatcase, a moretargetedpolicymightbe geared
core of the paper,in which the dataandresultsare towardstudentassistance,as opposedto reducing
describedin sequence.Finally,the studyis summa- geographicbarriersin a physicalsense.
rized in the conclusion.
Of course, there are instanceswhen additional
spacesareneededto meetchangesin demand.Ques-
POLICY
CONTEXT tions on how to allocatethese spacesarise,as well
as how these decisions may impact local access.
The findings in this study will certainly inform Theseissues playeda role in the debatesleadingup
policy debatesconcerningpostsecondaryfunding. to the Ontario"doublecohort"yearof 2003-04. In
One topic thatarisesin these debatesis the relative essence, the OntariogovernmenteliminatedGrade
merits of the different channels through which 13 fromthe high-schoolcurriculum,whichwas re-
policymakerscan increasepostsecondaryfunding quiredfor admittingOntariostudentsinto Ontario
to reduceaccess gaps.Geographicaccess gapsmay universities.Thelast Grade13 studentswereslated

CANADIANPUBLICPOLICY- ANALYSEDE POLITIQUES,


VOL.XXX,NO. 4 2004
430 MarcFrenette

to graduatein 2002-03 along with Grade 12 stu- (including family backgroundcharacteristicsand


dents,forminga doublecohortof studentsentering geography).The Surveyof LabourandIncomeDy-
the marketfor universityand college admission. namics (SLID) satisfies these requirementsquite
Additionalspaceswere necessaryto meet the sud- effectively. SLID is a longitudinalhouseholdsur-
demand,butwhichschools
denrisein postsecondary vey using the LabourForce Survey (LFS) as a
shouldhavereceivedthe additionalfunding? samplingframe.Eachpanelin SLIDis interviewed
for up to six years,anda new (overlapping)panelis
In the 2002 Ontariobudget speech, it was an- introducedeverythreeyears.Thefirstpanelstarted
nounced that postsecondary funding would be in 1993 and is now completeup to 1998. The sec-
increasedby $75 millionto accountfor reviseden- ondpanelbeganin 1996andcurrentlyhasfouryears
rolmentprojectionsin the doublecohortperiod.In of data(up to 1999).The mostimportantfeatureof
the samespeech,it was statedthat"we also recog- this datasetis thatthe postal code of the students'
nize thatnorthernandruralcollegesanduniversities home while in high school is available.
face financialchallengesdue to economiesof scale
and geography." As a result,it was announcedthat Which students are at risk of attending a
an additional$16 millionin operatingfundswould postsecondaryinstitution?In mostcases, one could
be allocatedto these institutions"to providefair- simplylookat studentsimmediatelyfollowinghigh-
ness andequity."4 school graduation(12 years of elementary and
secondaryschoolingin mostprovinces).In Quebec
TheOntario-based group"PeopleforEducation" and Ontario,however,universityentrancerequire-
expressedconcernoverthe possibilitythatstudents mentsaredifferent.Quebecstudentsmustcomplete
from largercities would have to face the prospect at leasttwoyearsof theuniversitystreamof CEGEP,
of attendingschools in distantareas.In a 2003 re- whereasOntariostudentsmustcompletetheirOn-
portentitledTheirFuturesat Stake,the groupcites tarioAcademicCredits(OACsor Grade13), at least
that northernand easterninstitutionswere able or duringthe periodof study.Since the objectiveis to
ready to take more studentsin the 2003-04 year look atthestudent'scircumstances whilelivingwith
whenexpressedas a percentageof theircurrentfirst- his or her family,it wouldbe inappropriate to ob-
yearenrolmentthanschoolsin southwestern Ontario serveQuebecstudentswhile in CEGEPsince some
andthe GreaterTorontoArea,whichmaybe indica- studentsleave home to attend,thus possiblyform-
tive of excess local demandin the morepopulous ing their own family.6 Similarto Frenette(2005),
areas.Theygo on to statethat"manydoublecohort the strategyappliedhere is to look at all students
studentswho would have gainedentranceto local who are two years awayfrombeing eligible to at-
institutionsin previousyearsmay not. Theywill be tenduniversityin theirhomeprovinceundernormal
forcedto considerdistantcolleges and universities. circumstances.7This requireslooking at students
Whetherdoublecohortstudentscan acceptthe addi- who just completedGrade 11 in Quebecand On-
tionalfinancialburdenof accommodation, andliving tario,andGrade10 in all otherprovinces(year"t").
expenses - on of
top high tuitioncosts - will be- Theirpostsecondaryparticipationpatterns(univer-
in
comea largerfactor postsecondary education."5 sity and college) are then observedover the next
two years (up to year "t+2").The highestlevel at-
tended is the one that predominatesin the final
DATA analysis (i.e., a studentwho attendscollege and
universitywouldbe coded as havingattendeduni-
The datarequirements of this studyareverysimilar versity).The sampleis furtherrestrictedto young
to Frenette(2005). First, longitudinaldata on stu- students- betweenthe ages of 15 and21 yearsold
dents "at risk"of attendinguniversityis required while in high school.

CANADIANPUBLICPOLICY- ANALYSEDE POLITIQUES,


VOL.XXX,NO. 4 2004
Accessto Collegeand University:
Does Distanceto SchoolMatter? 431

The second data requirementconsists of the Quebecstudents).Note thatstudentswho attended


postalcodesof Canadianpostsecondaryinstitutions college and universitywould be coded as having
andcolleges).This
(universities,university-colleges, attendeduniversity,althoughundernormalcircum-
informationis availablefrom the Web sites of the stances,one wouldonly expectthis to occurin the
Associationof UniversitiesandCollegesof Canada Quebecsystemwheretwo yearsof CEGEParere-
(AUCC)andthe Associationof CanadianCommu- quiredpriorto attendinguniversity.
nity Colleges (ACCC).8 The AUCC Web site
contains the postal codes of 101 publicly funded Distance to PostsecondaryInstitutions:a series
degree-granting universitiesanduniversity-colleges of dummyvariablesindicatingwhethera university
(includingall campuses),while theACCCWebsite anda college arewithincommutingdistanceof the
containsthe postal codes of 377 "communitycol- student'shome while in high school (the omitted
leges" (including all campuses).9 The term category),only a college is withincommutingdis-
"communitycollege"refersto publiclyfundednon- tance,andneithera universitynora collegeis within
university postsecondaryinstitutions, excluding commutingdistance.
trade-vocationalschools and business colleges.10
Many provincesrefer to them as communitycol- FamilyIncome:dummyvariablesindicatingthe
leges, buttheyareknownas CEGEPsin Quebecand income tier (meaningone-thirdin this context)of
as colleges of applied arts and technology in the student'seconomicfamilywhile in high school.
Ontario.1 Familyincomesareclassifiedby tierswithinthefive
standardareasizes of residencein orderto (partially)
The nextstepis to calculatethe distancebetween accountfor differencesin the cost of living, as well
thestudent's
homepriortograduating fromhighschool as the family'srelativesocio-economicstatusin the
andthenearestpostsecondary whichis done
institution, community.14The five sizes include rural;small
with the postalcodes of the studentandthe institu- urban(under30,000people);30,000-99,999people;
tions.TheAppendixdescribesthisprocessin detail. 100,000-499,999 people; and 500,000 or more
people. The middle-incometier is the omittedcat-
Threegroupsof studentsare of interestfor this egory.Note thatthe incomeis adjustedfor the size
study:those living withincommutingdistanceof a of the familyin orderto createa per capitaincome
universityand a college, those living withincom- measurethataccountsfor economiesof scale asso-
mutingdistanceof a college only, andthose living ciatedwith largerfamilies.The preciseadjustment
beyond commutingdistanceof a universityand a consistsof dividingfamilyincomeby thesquareroot
college.12 Studentsliving beyond 80 km from a of the size of the family.
postsecondaryinstitutionare classified as living
beyondcommutingdistance.13Note thatthisrefers ParentalEducation:dummyvariablesindicating
to the straight-linedistancebetweenthe two points, the highest level of educationalattainmentof the
and may correspondto a longerdrivingdistance. parents(the highestlevel attainedby eitherparent
is a universitydegree, a college diploma, or no
The set of variablesusedin the analysisincludes: postsecondarycertificate/don'tknow).

Postsecondary Attendance:a categoricalvariable Female:a dummyvariableto accountfor differ-


no
indicating postsecondaryattendance(0), college ences in postsecondaryparticipationratesbetween
attendance(1), or universityattendance(2) shortly the sexes.
afterhigh school (up to year"t+2,"the firstyearof
college/universityeligibility for non-Quebecstu- Province:a series of dummyvariablesindicat-
dents,andthe first yearof universityeligibilityfor ing the provincethe studentlived in while in high

CANADIANPUBLICPOLICY- ANALYSEDEPOLITIQUES,
VOL.XXX,NO. 4 2004
432 Marc Frenette

school, with Ontarioas the omittedcategory.This attrition).The sampleoverrepresentsthe year 1998


can captureinherentdifferencesin postsecondary (regardingpostsecondaryparticipation)since the
participationacrossprovinces,eitherdue to differ- SLID panels startoverlappingin 1996 (the high-
ences in student composition, differences in school students are sampled in 1996, and their
economicconditionsacrossprovinces,differences postsecondary upto 1998is observed).
participation
in tuition fees across provinces,or differencesin The sampledropsagainin 1999sincethefirstpanel
the academicrequirementsfor college and univer- endedin 1998. The samplingweights aredesigned
sity admittance forQuebecand
(especiallyimportant to adjustfor this asymmetry.
Ontariostudents).

Year:a seriesof dummyvariablesindicatingthe


TABLE1
yearup to whichwe observepostsecondaryattend- Sample Means
ance (year "t+2").This can capturetrendsin other
factorsthatmay affectpostsecondaryparticipation education = university 0.192
levelattended
Highest
(e.g., improvingeconomicconditionsor risingtui- education
Highest = college
levelattended 0.200
tion fees in the 1990s). Topincome
tier 0.403
Middleincome tier 0.335
The samplemeansof these variablesappearbe- Bottom incometier 0.263
low in Table 1. An equal proportionof students
and
University collegenearby 0.831
Collegenearby 0.135
attendeduniversityand college shortlyafterhigh Nouniversity/college 0.034
nearby
school (aboutone in five attendedeach).Note that Highesteducationallevelofeither
some studentswho attendeduniversitymay have parent= university
degree 0.158
also attendedcollege, butthe highestlevel attended Highesteducationallevelofeither
About40 percentof the samplecomes parent=collegediploma 0.242
predominates. Neither hasa postsecondary
fromfamiliesin the top incometier (withinan area parent
certificate
ordon'tknow 0.601
of a givensize), butthis is dueto the impliedage of Female 0.475
the parentsin the sample relative to the general Newfoundland 0.029
population- by design,theyhaveat leastone child Prince
Edward Island 0.005*
who is between the ages of 15 and 21 years old. NovaScotia 0.039
NewBrunswick 0.034
Most studentsare withincommutingdistance(less
Quebec 0.116
than80 km) of a universityand a college (83 per- Ontario 0.431
cent). Another13.5 percentare withincommuting Manitoba 0.043
distancefroma college only,while only 3.4 percent Saskatchewan 0.046
are out-of-commutingdistancefromeithertype of Alberta 0.108
institution.Recall that studentswho were within British
Columbia 0.148
1995 0.114
commutingdistance from a university,but not a 1996 0.185
college, areexcludedfromthe studysince thereare 1997 0.158
veryfew of thesecases.Thesampleunderrepresents 1998 0.359
Quebecstudentssimplybecausefamilyinformation 1999 0.186
hadto be garneredby lookingbackone year(since
N 2,065
someGrade11 studentsmayhavealreadyleft home
to attendCEGEPor to work).This eliminatedthe
Note:Inthisandallfollowing theterm"nearby"
tables,
first wave of the first two panels (1993 and 1996, tolessthan80kmofstraight-line
refers distance.
respectively),as well as anynewcross-sectionaltop- * Estimate
shouldbeviewedwithcaution.
ups (whichare addedto the sampleto accountfor

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Does Distanceto SchoolMatter?
Accessto Collegeand University: 433

RESULTS province, with Saskatchewanrankingthe lowest,


followedby Newfoundland andLabrador.15 Wenow
The GeographicProximityto Collegesand see thatthemajorityof thepopulationin Newfound-
Universities landandLabrador andSaskatchewan is nevertheless
Table 1 answeredthe question "Whereare high- withincommutingdistanceto a universityor a col-
schoolstudentslocatedrelativeto thenearestcollege lege, as many of their residents are within
and/oruniversity?"Due to smallprovincialsample commutingdistanceto a college only. At least 89
sizes, this questioncould not be investigatedat the percentof residentsin anyprovincelive withincom-
provinciallevel. However,by using the 1996 SLID mutingdistanceof a universityor a college (see the
cross-sectionalsample, the location of the entire TotalcolumnunderUniversityor CollegeNearby).
populationrelativeto universities/collegescan be Rankinghighest are Prince EdwardIsland, Nova
ascertained.Table2 shows the distributionof the Scotia,andNew Brunswick(100 percenteach),fol-
populationby distanceto thenearestuniversityand/ lowed by Ontario(99 percent).Rankinglowest is
or college for all Canadiansandby province. Manitoba(89 percent),followedby Newfoundland
and LabradorandAlberta(93 percent).In general,
First, note that the nationalproportionsfor all the vast majorityof the Canadianpopulationlives
Canadians(Table2) are quiteclose to the national neara universityor a college. Most of the analysis
proportionsfor all students (Table 1). Frenette will thusfocus on comparingthe participationrates
(2005) found that there was tremendousvariation of studentswho live neara universityanda college
in the proximityof universitiesto Canadiansby to those of studentswho live neara college only.

2
TABLE
Distribution in1996byDistance
of Population to NearestUniversity
and/orCollege

Type ofInstitution
Nearby
or Nearby
UniversityCollege Neither
Province and
Sample University nor Total
University
College CollegeOnly Total College
Nearby

Newfoundland
andLabrador 3,588 0.579 0.347 0.926 0.074 1.000
PrinceEdward
Island 1,875 0.907 0.093 1.000 0.000 1.000
NovaScotia 4,790 0.957 0.043 0.999 0.001 1.000
NewBrunswick 4,535 0.858 0.142 1.000 0.000 1.000
Quebec 13,470 0.905 0.064 0.969 0.031 1.000
Ontario 20,462 0.921 0.073 0.993 0.007 1.000
Manitoba 4,845 0.789 0.106 0.895 0.105 1.000
Saskatchewan 4,729 0.488 0.478 0.965 0.035 1.000
Alberta 6,069 0.771 0.162 0.933 0.067 1.000
British
Columbia 6,367 0.848 0.124 0.971 0.029 1.000
Canada 70,730 0.867 0.106 0.973 0.027 1.000

Note:Excludes
thosewithonlya university
nearby.

CANADIANPUBLICPOLICY- ANALYSEDE POLITIOUES.


VOL.XXX.NO. 4 2004
434 Marc Frenette

Distanceto Schooland Postsecondary Anotherinterestingresult from Table 3 is that


Attendance studentsliving beyondcommutingdistancefroma
collegearefarless likelyto attendcollegethanthose
DescriptiveAnalysis living withincommutingdistance.Recall,however,
Looking first at the raw data, Table 3 shows the thatonly 3.4 percentof studentslive beyondcom-
postsecondaryattendanceratesby the typeof insti- mutingdistancefrom a college (Table 1), so it is
tution nearby(within 80 km). Studentswithin 80 perhapsless of an issue thandistanceto university.
kmof a universityanda college areequallyas likely
to attenduniversityor college (aboutone in five, or EconometricAnalysis
20 percent,attendeach type of institution).When In this section, the role of distance to school in
only a college is within80 km, the universitypar- postsecondary participationis analyzedmoreclosely
ticipationrate falls to 13 percent,and the college in a multinomiallogit model with threeoutcomes:
participationraterises to about22 percent.On the no postsecondary schooling,collegeattendance,and
surface,it appearsthat the uptakein college par- universityattendance.17
ticipationis quite low comparedto the decline in
universityparticipation;however,these results do Model 1: The OverallRole of Distance. Students
not accountfor differencesin observablecharacter- who live beyondcommutingdistancefrom a uni-
istics. As it turnsout, Quebecstudentsplay a large versityareless likely to attenduniversity(Frenette
role in these results for two reasons:they are far 2005). If these studentsare withincommutingdis-
morelikelyto attendcollege(CEGEP),16 anda rela- tance to a college, do they "make up" for this
tively highproportion of themlive neara university disadvantage by beingmorelikelyto attendcollege?
and a college (Table2). By excludingQuebecstu-
dents,theuptakein collegeparticipation in outlying The explanatoryvariablesincludedin the first
areas almost fully counterbalancesthe decline in model are distanceto school, family income, pa-
universityparticipation(see the bottomof Table3). rental education, sex, province, and the year by
In the regressionanalysis,Quebecstudentswill be which we observe whether the student attended
included,but the provinceof residencewill serve postsecondaryschooling or not. The results are
as a controlvariable. shownbelow in Table4.

3
TABLE
andNon-University
University Participation
Postsecondary to NearestUniversity/College
byDistance

Proportion
Attending
ofInstitution
Type Nearby Sample University orCollege
College University

Allstudents
andcollege
University 1,489 0.205 0.200 0.404
Collegeonly 448 0.130 0.217 0.347
Neither 128 0.112 0.134 0.246

Quebecstudentsexcluded
andcollege
University 1,346 0.229 0.146 0.376
Collegeonly 430 0.136 0.204 0.340
Neither 113 0.136 0.070 0.205

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Accessto Collegeand University:
Does Distanceto SchoolMatter? 435

4
TABLE
Attendance
Postsecondary 1 (multinomial
Model logit)

DependentVariable
0 = nouniversity/college
attendance
1= college
attendance Attendance
University Attendance
College
2 = university
attendance (PS=2) (PS= 1)

Intercept -1.990 *** -1.470 **


(-4.59) (-4.76)
Collegenearby -0.526 *** 0.319*
(-2.64) (1.65)
Nouniversity/college
nearby -0.539 -0.707**
(-1.64) (-1.98)
Topincometier 0.360* 0.209
(1.77) (1.07)
Bottomincometier -0.672 *** -0.056
(-2.64) (-0.24)
levelofeither
educational
Highest is a university
parent degree 1.362*** 0.110
(5.49) (0.46)
levelofeither
educational
Highest is a college
parent diploma 0.371 -0.186
(1.53) (-0.85)
Female 0.576 *** 0.208
(3.2) (1.24)
Newfoundland 0.507* -0.676 **
(1.87) (-2.19)
Edward
Prince Island 0.561 -0.944*
(1.53) (-1.69)
NovaScotia 0.426 -0.878 **
(1.51) (-2.37)
NewBrunswick 0.278 -0.382
(1.07) (-1.25)
Quebec -1.559 *** 1.639***
(-3.1) (6.13)
Manitoba 0.093 -0.649*
(0.31) (-1.76)
Saskatchewan -0.003 -0.679*
(-0.01) (-1.93)
Alberta -0.546 * -0.563 *
(-1.91) (-2.06)
British
Columbia -0.812 ** -0.128
(-2.49) (-0.48)
1996 0.845** 0.119
(1.96) (0.37)
1997 0.648 -0.226
(1.58) (-0.66)
1998 0.329 0.206
(0.9) (0.78)
1999 0.084 0.068
(0.22) (0.23)
N 2,065
ratio)
-2*log(likelihood 3,432.65

***significant
Notes:z statisticsinparentheses; **significant
at 1 percent; * significant
at 5 percent; at 1 percent.

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


436 Marc Frenette

Relativeto studentswith a universityanda col- areto attenduniversityor college.Oneprovincethat


lege nearby,studentswithonly a collegenearbyand standsoutis Quebec.Comparedto Ontario,students
thosewith neithera college nora universitynearby in Quebec are far less likely to attenduniversity,
are less likely to attenduniversity(by roughlythe butarefarmorelikely to attendcollege. This might
samemargin,given the equalmagnitudeof the co- be partlybecausemanyQuebecstudentsarestill in
efficients).Studentswith a college nearbyaremore CEGEPtwo yearsaftercompletingGrade11, even
likelyto attendcollege thanarestudentswitha uni- if they plan on pursuinga universityeducation.A
versity and a college nearby(with a significance longertime framewouldbe necessaryto moreac-
level of 10 percent).Thus,thereis an uptakein col- curatelydepictthesituationin Quebec,butthefocus
lege attendanceamongstudentsliving away from of this paperis on the nationalresults(usingprov-
universities,butwho areneverthelessnearcolleges. ince as one of manycontrolvariables).
Studentswho live awayfromcolleges, on the other
hand,areless likely to attendcollege thanstudents Model 2: Interact Family Income with Distance
who are near a college, whether a university is to School. Since we know that the studentsmost
nearbyor not (bothsignificantat 5 percent).18 likely to be disadvantagedin accessinguniversity
as a result of commutingdistanceare those from
Relativeto the middle-incometier, studentsin lower-incomefamilies(Frenette2005), it wouldbe
the topincometieraremorelikely to attenduniver- instructiveto knowif these studentsaretakingad-
sity,while studentsin the bottomtier areless likely vantageof theirproximityto college (to make up
to attend.The magnitudeof the coefficienton the for their disadvantagein accessinguniversity).To
bottomtier variableis not only largerthanthatof answerthis question,the distancevariablesarein-
the top tier (in absolutevalue),it is also significant teractedwiththeincomevariablesin model2, which
at 1 percent(comparedto 10percentforthetoptier). is shownbelow in Table5.
In terms of college attendance,income does not
seem to play a strongrole, as the coefficientsare To focus on the questionof college uptakein
smallin magnitudeandnot statisticallysignificant. outlyingareasby incometier, we need to examine
the college attendancecoefficients. More specifi-
Parentalpossessionof a universitydegreeis as- cally, we need to look at the income/distance
sociated with a higher probabilityof university interactioncoefficients.Forthetopincometier,stu-
participation,but havinga college-educatedparent dentsneara collegeonly areno morelikelyto attend
only has a small positiveimpacton the likelihood college than studentsnear a universityand a col-
of attendinguniversity (relative to not having a lege (thecoefficienton the"topincometier*college
postsecondary-educated parent).Forcollege attend- nearby"variableis slightly negative).The uptake
ance,parentaleducationseemsto matterverylittle. in college attendanceappearsto occur amongstu-
dentsfromlower-andmiddle-incomefamilies.Both
Consistent with published statistics from the the "bottomincome tier*collegenearby"and the
1990s,femalesare morelikely to attenduniversity "middleincome tier*collegenearby"are positive,
thanmales(significantat 1 percent).19Femalesare althoughnotquitesignificantatthe 10percentlevel.
only slightly more likely to attend college than
males, however,and the differenceis not statisti- For all threeclasses of income, studentsliving
cally significant. far awayfromcolleges areless likely to attendcol-
lege than students living near both types of
Manyprovincialcoefficientsare not significant, institutions.Recallfrommodel1 thatthecoefficient
indicatingthatstudentsin theseprovincesareabout is significantat 5 percentwhen it is not interacted
as likelyas Ontariostudents(thereferencecategory) with family income. Note that the samplesize of

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Accessto Collegeand University:
Does Distanceto SchoolMatter? 437

TABLE5
Postsecondary Model2: Distance
Attendance andIncomeInteracted
(multinomial
logit)

Dependent Variable
0 = nouniversity/college
attendance
1 = collegeattendance Attendance
University Attendance
College
2 = university
attendance (PS= 2) (PS= 1)
Intercept -1.997 *** -1.510 ***
(-4.57) (-4.72)
tier
Topincome 0.346 0.287
(1.51) (1.28)
Bottomincometier -0.571 ** -0.068
(-2.05) (-0.24)
Topincome
tier*college
nearby -0.401 -0.096
(-1.3) (-0.31)
Middle
income
tier*college
nearby -0.358 0.501
(-1.27) (1.59)
Bottomincometier*collegenearby -1.601 *** 0.568
(-3.28) (1.55)
tier*no
Topincome university/college
nearby -0.213 -0.587
(-0.53) (-1.06)
Middle
income
tier*no
university/college
nearby -1.175 * -0.656
(-1.87) (-1.18)
Bottom
income
tier*no
university/college
nearby -0.521 -1.023
(-0.5) (-1.27)
Highest levelofeitherparent
educational is a university
degree 1.377*** 0.101
(5.52) (0.42)
Highest levelofeitherparent
educational is a collegediploma 0.379 -0.179
(1.57) (-0.82)
Female 0.585 *** 0.212
(3.25) (1.26)
Newfoundland 0.504 * -0.703 **
(1.84) (-2.26)
Prince
Edward
Island 0.546 -0.948 *
(1.5) (-1.69)
NovaScotia 0.410 -0.874 **
(1.46) (-2.36)
NewBrunswick 0.249 -0.376
(0.96) (-1.22)
Quebec -1.568 *** 1.646 ***
(-3.11) (6.1)
Manitoba 0.073 -0.628 *
(0.24) (-1.7)
Saskatchewan 0.001 -0.696 **
(0) (-2.04)
Alberta -0.574 ** -0.562 **
(-1.99) (-2.04)
BritishColumbia -0.833 ** -0.115
(-2.55) (-0.43)
1996 0.846 ** 0.124
(1.97) (0.38)
1997 0.648 -0.215
(1.58) (-0.63)
1998 0.329 0.209
(0.9) (0.79)
1999 0.071 0.079
(0.19) (0.27)
N 2,065
-2*log(likelihood
ratio) 3,423.72

***significant
Notes:z statisticsin parentheses; at 1 percent;** significant
at 5 percent;* significant
at 1 percent.

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


438 MarcFrenette

eachclass of incomein thisdistancecategoryis very abilityof participatingin anytype of postsecondary


low, whichmightbe one reasonwhy the resultsare schoolingis also shown.This is simplythe sum of
not significant.Anotherreasonis thatthe reference the universityand college predictedprobabilities.
groupconsistsof studentsneara college anda uni- All probabilitiesin Figure 1 arebasedon the coef-
versity, many of whom choose university over ficientsestimatedin the first model.
college. Studentsfrom lower- and middle-income
families are less likely to attendcollege when one Whena universityanda college are nearby,stu-
is not nearbythanwhenone is nearby(buta univer- dents are as likely to attendeithertype of school.
sity is not nearby),with a significancelevel of 10 When only a college is nearby,the universitypar-
percent. ticipationrate falls considerably,but the college
participationrate increases substantially.On bal-
TheMagnitudeof the Role of Distance to ance, the overall postsecondaryparticipationrate
School (universityandcollege combined)is aboutthe same
So far, we have examinedthe role of distanceto for both types of students - about 40 percent.
school in a series of multinomiallogit models.To Hence, despite the fact that studentswho live be-
betterappreciatethe magnitudeof this role, pre- yondcommutingdistancefroma universityareless
dictedprobabilitiesof postsecondaryparticipation likely to attenduniversity,they arejust as likely to
were generatedfromthe regressions.20 pursuepostsecondaryschooling in general - as
long as a college is nearby.Whetherthis is the first
Figure1 showsthepredictedprobabilitiesof uni- choice of thesestudentsis obviouslya questionthat
versityandcollege participationfor studentsliving cannotbe answeredwithinthe scope of the dataat
in thethreedistancecategories.Thepredictedprob- hand.

1
FIGURE
Predicted
Postsecondary RatesbyDistance
Participation toSchool

0.50
0 UniversityParticipation
0.45 College Participation
0.40 0 Universityor College Participation
0.35
0.30
0.25
0.20
0.15
0.10 -
0.05
0.00
Universityand College Nearby College Nearby No University/CollegeNearby

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Accessto Collegeand University:
Does Distanceto SchoolMatter? 439

Forthe groupof studentswho live beyondcom- Figure2 shows the postsecondaryparticipation


muting distance from a college, the college rates for the three income tiers, which are gener-
participationrate is 37 percentlower than that of ated from model 2. We see that for studentsfrom
studentsliving withincommutingdistance.21This families in the top income tier, the college-partici-
suggeststhatdistancepresentsa challengeto some pation rate is the same whether they are near a
studentscontemplatinga college education.Never- universityanda college, orjustneara college. Since
theless,only3.4 percentof studentsin thesamplelive the universityparticipationis lower for these stu-
farawayfroma college.As outlinedinFrenette(2005), dents when they are beyond commutingdistance,
distanceto school presentsa challengeto far more we also see a moderatedeclinein theiroverallpar-
potentialuniversitystudents:aboutone in five. ticipationrate.Studentsfromfamiliesin themiddle
andbottomincometiers, on the otherhand,arefar
theuniversityparticipation
Interestingly, ratealso more likely to attendcollege when living near a
increaseswhena college is no longerwithina local college only(relativeto studentsin thesameincome
commute.Thismightresultfromthe fact thatsome tier andliving neara universityand a college). The
studentswouldhave chosen college if thatwas the overallpostsecondaryparticipationrateis aboutthe
only local choice; when no institution(college or sameforstudentsin lower-andmiddle-income fami-
university)is nearby,some of these studentsmay lies whetherthey are near a
just college or nearboth
chooseto attenduniversity(sincetheyhaveto leave a universityanda college.
home regardless).

2
FIGURE
Predicted
Postsecondary RatesbyDistance
Participation toSchool

0 University
Participation
0.50
TopIncomeTier
1-.
MiddleIncomeTier
CollegeParticipation
D University
or CollegeParticipation
0.45
BottomIncomeTier
0.40

0.35

0.30

0.25 -

0.20

0.15

0.10

0.05
0.00
Univ./Coll. Coll. No Univ./Coll. UnivJColl. Coll. No Univ./Coll. UnivJColl. Coll. No UnivJColl.
Nearby Nearby Nearby Nearby Nearby Nearby Nearby Nearby Nearby

CANADIANPUBLICPOLICY- ANALYSEDE POLITIQUES,


VOL.XXX,NO. 4 2004
440 MarcFrenette

Collegeattendanceis lowerfor studentswho live families.Nevertheless,distanceto schoolis gener-


beyondcommutingdistancefroma college regard- ally less of an issue for colleges than it is for
less of theirfamilyincome,butthis negativeeffect universities.Only3 percentof studentslive beyond
is felt far moreamongstudentsfromlower-income commutingdistanceto a college, while one in five
families.RecallfromTable5, however,thatthe co- students live beyond commuting distance to a
efficients generating these differences are not university.
statisticallysignificant.
The patternsof universityandcollege participa-
tionby familyincomeanddistanceto school(Figure
CONCLUSION 2) are consistentwith the notion that addedcosts
deterstudentsin less favourableeconomiccircum-
Previousworkhas foundthathigh-schoolstudents stances from pursuing a university or college
living beyondcommutingdistanceto universityare education.When no universityis nearby,students
farless likelyto attendshortlyafterhighschoolthan from lower-incomefamilies are far less likely to
those living withincommutingdistance,especially attendthan studentsfrom upper-incomefamilies.
amongstudentsfrom lower-incomefamilies. This The same can be said when no college is nearby,
study seeks to answerthree follow-up questions. althoughthis situationis far less common.In both
First, are studentswho live far from universities cases, however,the impacton studentsfrom mid-
morelikely to attendcollege if one is nearby,even dle-incomefamilieslies somewherein betweenthe
if this may or may not be their first choice? And impacton studentsfrom lower- and upper-income
given thatlower-incomestudentsare so negatively families.
affected by distancein accessing universities,do
they "makeup"for this disadvantageby attending As mentionedin thepolicycontextsection,these
college? Finally, does distance to college pose a resultsinformdebatesregardinguniversityfinanc-
substantialbarrierto studentscontemplatinga col- ing.Addressinggeographicaccessissuesis normally
lege education? undertakenthroughtwo avenues:increasedassist-
ance to students and increased funding to
The resultsindicatethat studentsliving beyond institutions.Since the geographiclocation of stu-
commutingdistancefroma universityarefar more dents can be easily identified,the study findings
likely to attenda college, as long as one is nearby. providepertinentinformationthat could assist the
As a result, postsecondaryparticipationrates are designof studentloan programs.The findingsalso
aboutthe samefor studentsliving neara college or informdebateson how to allocatenew fundsacross
botha college anda university- about40 percent. universitiessince such decisionsoften have impli-
Thisuptakein collegeattendanceoccursamongstu- cationsfor local access.
dents from lower- and middle-income families
only - those most negativelyaffectedby distance
in accessing universities. Students from upper- NOTES
incomefamiliesareno morelikely to attendcollege
Anyviewsthatmaybe expressed in thispaperarethose
whenthis is the only choice nearby. of theauthoranddonotnecessarilyreflectthoseof Sta-
tisticsCanada.Helpfulcommentswerereceivedfrom
Studentswho live beyond commutingdistance GeorgeButlin,LouisChristofides, GarnettPicot,and
from a college are 37 percentless likely to attend at
participants theCanadian
Economics meet-
Association
than those who live closer. As with universityat- ings at CarletonUniversity,May 2003, and at the
tendance,the negativeeffect of distanceto college Canadian Millennium
ScholarshipConference on Path-
is felt far moreamongstudentsfromlower-income ways to Access, Ottawa,October2003. RussellWilkins

CANADIANPUBLICPOLICY- ANALYSEDEPOLITIQUES,
VOL.XXX,NO. 4 2004
Accessto Collegeand University:
Does Distanceto SchoolMatter? 441

was veryhelpfulin providingadviceon usingthe PCCF+ cline in thenumberof colleges"atrisk"of beingattended.


in this andearlierwork.All remainingerrorsare the re- Note thatin Frenette(2005), universityparticipationpat-
sponsibilityof the author. ternswererobustto theinclusionof all universities.Thus,
the currentstudyincludesall colleges anduniversitiesin
'Foran overviewof recentdevelopmentsin the litera-
the analysis.
ture,see Card(1999).
12A very tiny proportionof studentslive near a uni-
2Forexample,there may be substantialcost savings
versity, but not near a college. These cases are omitted
associatedwith buyinglargerfood items. If the student
since no substantialanalysiscouldbe performedon them.
leaves home to attendschool, the two householdsmay
have to separatelypurchasesmallerfood items than if 13Frenette (2005) had threedistancecategories:0-40
the studenthadremainedat home(to avoidspoilage).Of km (within commutingdistance), 40-80 km (possibly
course,the studentmayrecoversome of these economies beyondcommutingdistance),and80 kmor more(beyond
of scale by sharinga dwellingwith otherstudents,so that commutingdistance).The smallercell sizes in the cur-
the student'sfamilymaysee the majorityof the foregone rent studynecessitatedcombiningthe 0-40 km and the
economiesof scale. 40-80 km categories;otherwise,nine categorieswould
need to be created.Note thatotherthresholdswere also
3See Christofides,Cirello and Hoy (2001) for an ex- used(e.g., 60km, 100km,etc.),buttheresultswerequali-
ampleof a studyusing tuitionas an explanatoryvariable
tatively very similarin each case. The 80 km threshold
in postsecondaryaccess. In a varietyof specifications,
was chosen since it is a distancebeyondwhichit would
the authorsdo not find a negativeand significanteffect.
be extremelydifficultto commute,especiallygiven that
4Thefull speechis availableat http://www.gov.on.ca/ this is a straight-linedistanceandmaycorrespondto sub-
FIN/bud02e/statement-eng.htm. stantiallygreaterdrivingdistance.

5The report is available via email at p4e@ '4Theresultsare robustto calculatingthe family in-
peopleforeducation.com. come tier acrossall of Canada.

6InSLID,this dependson whetheror nottheyreturned '5Note that Frenette(2005) includedonly 71 of the


to live with their parentsduringthe year.If they return 101 universities.This studyincludesall 101 universities.
for at least 30 days, say in the summer,they are still Theresultsin Frenette(2005)arerobustto thisrestriction.
groupedtogetherwiththeirparents;otherwise,theyform
16ManyQuebecstudentsarestill in CEGEPtwo years
theirown family.
afterGrade11, even if they eventuallyplanon attending
7Studentsare eligible to attendcollege no later than university.As Frenette(2005) points out, a longertime
when they are eligible to attenduniversity. horizonwouldbe necessaryto moreaccuratelydepictthe
particularsituationin Quebec.
8TheWeb sites can be visited at http://www.aucc.ca
and http://www.accc.ca,respectively. 17Thegeneralmodelestimatedis Ln =
[Pij/(Pio)]Xi3j"
+i,
where"P"is the probabilityof postsecondaryparticipa-
9Thelist of universitiesandcolleges is availableupon tion, "x"is a vector of regressors,and "?"is a random
request. disturbanceterm.The subscript"i"denotesthe student,
while the subscript"j"denotesthe typeof postsecondary
'0lnformationon the postalcodes of tradeschoolsand
businesscolleges are not availableto the author. schooling(0 = no postsecondary;
1 = college;2 = university).
The referenceoutcomeis "nopostsecondaryschooling"
"In Frenette(2005), only universitiesof interestto (j= 0).
the generalstudentpopulationwere included(71 of the
'8Onemightwonderif the distanceto school effect is
original 101, which accountedfor about91 percentof
the entirestudentbody).A non-negligibleproportionof simply a "ruraleffect,"insomuchas ruralstudentsmay
simplybe moreinclinedto attendcollege as opposedto
colleges aremainlygearedto thelocal labourmarket,and
thusofferprogramsin a veryspecificrangeof disciplines. university,and distancemay be highly correlatedwith
ruralstatus.To test this notion,an additionalregression
Eliminatingthese colleges wouldhave led to a largede-

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


442 Marc Frenette

was estimatedwherebythe distancevariableswereinter- Thepredictionsarecalculatedat theindividuallevel, and


acted with a rural indicator variable. Among rural then averagedout overthe entiresample(basedon vari-
students,distanceplays the exact samerole as described ous scenariosof interest). See Frenette(2005) for the
by model2 (i.e., comparedto those with local access to advantagesof this technique.
both a college anda university,those with only local ac-
cess to a college are far less likely to attenduniversity, 21Thecollege participation rateamongall studentsliv-
ing nearcolleges had to be calculated
to obtainthis figure.
butarefarmorelikelyto attenda college instead).Among
To this end, it was necessaryto calculatea population-
urbanstudents,we also see a solid decline in university
attendance(when only a college is nearby),but a more weightedaverageof the predictedcollege participation
rates among studentsliving near both types of institu-
moderateincrease in college attendance.In summary,
tions andthose living neara college only.
additional distance to school appearsto matter even
amongruralstudents,and(to a lesser extent)amongur-
ban students.Specificresultsareavailableuponrequest. REFERENCES
Onemightalso wonderaboutthepossibilitythatsome
studentsmaypreferto attendcollege for a yearor two in Andres,L. and E.D. Looker.2001. "Ruralityand Capi-
orderto save costs, andthentransferto a university.This tal: Educational Expectations and Attainmentsof
can be a concernin BritishColumbiaandAlbertain par- Rural/UrbanandMetropolitan Youth,"The Canadian
ticular, where some colleges have university transfer Journalof HigherEducation31(2):1-46.
credits.In essence, the fact thatstudentsare morelikely Butlin,G. 1999. "Determinants of PostsecondaryEduca-
to attendcollege when thatis the only local optionmay tion,"EducationQuarterlyReview5(3):9-35.
be largelyexplainedby universitytransfercredits,at least Card,D. 1999. "TheCausalEffectof Educationon Earn-
in these provinces.If a sufficientnumberof studentsin ings," in Handbookof LaborEconomics,Volume3,
thesecollegeseventuallymakethetransfer,thenwe would ed. O. Ashenfelterand D. Card.Amsterdam:North-
expect to see a dampeningof the distanceeffects as we Holland.
look furtherinto the student'spostsecondaryschooling Christofides,L., J. Cirello and M. Hoy. 2001. "Family
years. To test this hypothesisat the nationallevel (pro- IncomeandPostsecondaryEducationin Canada,"The
vincial sample sizes are simply too small), the time CanadianJournalof HigherEducation31(1):177-208.
horizonof the estimationswereextendedto the student's Corak,M., G. Lipps and J. Zhao. 2003. FamilyIncome
third year of college/universityeligibility, which still and Participationin PostsecondaryEducation.Ana-
showed very solid distance effects along the lines de- lyticalStudiesResearchPaperSeriesNo. 210. Ottawa:
scribed in Table 4. Specific results are availableupon StatisticsCanada.
request. Frenette,M. 2005. "TooFarto Go On?Distanceto School
and UniversityParticipation," EducationEconomics,
19SeeCANSIMtables 580602, 580603, 580701, and
forthcoming.
580702 for moredetailson universityenrolmenttrends.
Mehmet,0. 1978. WhoBenefitsfrom the OntarioUni-
20Thiswas done by applyingthe following general versitySystem?Toronto:OntarioEconomicCouncil.
formula:Pij= exp(xi'bj)/[1+ Ekexp(xi'bk)],k = 1 to 2 Meng,R. andJ. Sentence.1982."Canadian Universities:
WhoBenefitsandWhoPays?"TheCanadianJournal
(college =1; university=2), where"Pij"is the predicted
probabilityof an individual"i"attendingpostsecondary of HigherEducation12(3):45-58.
schooling "j",and "x,'b"is a linearcombinationof the Thiessen,V. and C. Nickerson.1978. CanadianGender
Trendsin Educationand Work.Ottawa:HumanRe-
regressorvariables(at a given set of values xi') each
sourcesDevelopmentCanada.
multipliedby theirestimatedregressioncoefficientin b.

CANADIAN PUBLIC POLICY - ANALYSE DE POLITIQUES, VOL. XXX, NO. 4 2004


Accessto Collegeand University:Does Distanceto SchoolMatter? 443

APPENDIX
TOSCHOOL
DISTANCE
CALCULATING

To calculatethe distancebetweenanytwo points,we simplyneedthe geographiccoordinatesof thepoints,


as well as some basic sphericalgeometry.The geographiccoordinates(latitudeandlongitude)of students
arederivedfromthe postalcodes of householdsby usingthe residentialversionof PostalCodeConversion
File Plus (PCCF+),a programthatconvertssix characterpostalcodes into variousgeographicunits,includ-
ing latitudeandlongitude.The institutions'geographiccoordinatesarecalculatedby usingthe institutional
versionof the PCCF+.Assumingthe earthto be a perfectspherewitha radiusof 6,370.997km,the formula
for the straight-linedistance(in kilometres)between the student'shome and the nearestpostsecondary
institutionis:

Distance= 6,370.997*arcos[sin(s_latrad)*sin(i_latrad)+cos(s_latrad)*cos(i_latrad)*cos(s_longrad-i_longrad)]

Thegeographiccoordinates(in degrees
Where"latrad"is the latitudein radians,andlikewisefor "longrad."
anddecimals)were convertedto radiansby dividingby 57.29577951.Note that"s_"denotesthe student's
locationand"i " denotesthe institution'slocation.

PUBLIC
CANADIAN - ANALYSE
POLICY DEPOLITIQUES,
VOL.XXX,NO.4 2004

También podría gustarte