the determinants of life satisfaction march 2016quality of life. housing, income and jobs are the...

The Determinants of Life Satisfaction

Philip Hoskins (Department of Finance, Government of Newfoundland and Labrador)

Douglas May (Memorial University, Canada)

Paper prepared for the 34

th IARIW General Conference

Dresden, Germany, August 21-27, 2016

Session 8F: Meeting the Measurement Challenges of Official Statistics Offices V

Time: Friday, August 26, 2016 [Afternoon]

PhilipHoskinsandProfessorDouglasMay

COLLABORATIONFORAPPLIED

RESEARCHINECONOMICS

THEDETERMINANTSOFLIFESATISFACTIONINNEWFOUNDLANDANDLABRADOR

PreparedfortheCollaborationforAppliedResearchinEconomics(CARE)initiative,DepartmentofEconomics,MemorialUniversityofNewfoundland

BytheCentrefortheStudyofLivingStandardsCAREResearchReport

Abstract

Happiness,subjectivewell-beingandutilityhavebeenatthecentreoftheneoclassicalmicroeconomicanalyticalframeworkforaconsiderableperiodoftime.Initsstrictestformtheutilitythatanindividualobtainsisdependentuponthecombinationofgoodsandservicesthatheorsheconsumes.Theactualorperceivedutilityofthesegoodsorservicesistheverycharacteristicthatmakestheindividualwanttoconsumethem.Atthesameithasbeenrecognizedthatthereareaspectsofourlivesbeyondconsumptionthathelptodetermineoursubjectivewell-being.Theseconcernsofmicro-economistshavespilledovertothoseinterestedinmacroeconomicsandtheSystemofNationalAccounts,whichfocusesontheproductionofgoodsandserviceswithinaneconomyorbyitsnationals.Forexample,in1974RichardEasterlinnotedthathappinessdoesnotseemtoincreasewithinacountrysuchastheUnitedStatesasGDPpercapitaincreases.Thisseemstoposeaparadox(calledthe”Easterlinparadox”)foreconomists.Overthepastdecadetherehavebeenanumberofstudiesthathaveattemptedtomeasurethedeterminantsofsubjectivewell-beingorlifesatisfactionandoursisoneofthem.Thesestudieshaveincreasinglyreliedonaccesstoappropriatecross-sectionaldataonindividualsandtomoresophisticatedeconometrictechniques.Ourstudyfollowsthattraditionbyusing2012-12CanadianCommunityHealthSurvey(CCHS)datafromStatisticsCanadaforNewfoundlandandLabradortoestimatethedeterminantsofsubjectivelifesatisfactionusingageneralizedorderedlogisticregressionmodelusingagologit3commandwrittenRichardWilliamsforuseintheSTATAstatisticalpackage.Ourresultsdemonstratetheaddedvalueofusingsuchaprocedureincludinganon-linearresponseofsubjectivelifesatisfactioncategoriestochangesinincomeastraditionaleconomictheorywouldpredict.Surprisingly,wealsofindthatgenderdoesnotmakeagreatdealofdifferenceinexplainingsubjectivewell-beingforoursample.

Inhisbook“Happiness:LessonsfromaNewScience”,theauthorRichardLayardwastesnotimeindescribingthemainproblemthatthis“newscienceofhappiness”ismeanttotackle.Intheveryfirstparagraphofchapteronehestates:

Thereisaparadoxattheheartofourlives.Mostpeoplewantmoreincomeandstriveforit.Yet

asWesternsocietieshavegotricher,theirpeoplehavebecomenohappier.

Wewillnotentangleourselvestoomuchhereinthedebateoverwhetherthisparadoxisrealornot.

Forourpurposes,wewillfocusonthetwounderlying“facts”thatsuggestthereisaparadoxatall.

Namely,whatdowemeanwhenwesaythatsocietieshavegottenricherandhowdoweknowthat

peoplehavebecomenohappier?

Ofcourse,whenwesaythatsocietieshavegottenricherwemeanthattheirGrossDomesticProduct

hasincreased.Theparadoxthatarichersocietydoesnotimplyahappiersocietyisoftenreferredtoas

the“Easterlinparadox”sinceitwasEasterlin(1974)whofirsthighlightedthefactthathappinessdoes

notnecessarilyincreasewithanincreaseinGDP.EasterlinanalyzedU.S.datafrom1946to1970and

concludedthateventhoughincomehadincreaseddramaticallyoverthattimeperiod,thelevelofself-

reportedhappinesshadremainedflat.Thisdoesnotimplythatmoneydoesnotbuyhappiness.Money

does“buy”happinessfortheindividualinthesensethat“atapointintimebothamongandwithin

countries,happinessandincomearepositivelycorrelated”(Easterlin1974).However,money,orgrowth

inGDP,doesnotnecessarilybuyoverallhappinessforacountry’scitizens.

EversinceitsdevelopmentbyAmericaneconomistSimonKuznetsinthe1930’s,theGDPhasbeenused

astheprimaryindicatorofanation’swealth.Butisitthebestmeasureofprogress?Inafamousspeech

attheUniversityofKansasonMarch18,1968RobertF.KennedyremarkedthatGDP“countsair

pollutionandcigaretteadvertising,andambulancestoclearourhighwaysofcarnage.Itcountsspecial

locksforourdoorsandthejailsforthepeoplewhobreakthem.”Allofthesenegativesaddtothe

growthofGDP.Itcountseverything,Kennedysays“exceptthatwhichmakeslifeworthwhile.”As

Anielski(2007)pointsouttheoriginal,OldEnglish,meaningofthewordwealthwas“thestate,

conditionorqualityofwell-being.”ThiswasprobablynotthemeaningofwordwealththatAdamSmith

wasreferringtowhenhewrotehistreatiseAnInquiryintotheNatureandCausesoftheWealthof

Nations,butperhapsitistime“gobeyondGDP”andproducemeasuresof“wealth”inthearchaicsense

ofthewordthatencompassbothmaterialandpsychologicalwell-being.Soonewaytoresolvethe

paradoxistoreplaceGDPwithamoreappropriatemeasureofhowrichasocietyis.

InthelastseveraldecadesresearchershaveendeavoredtoproducesuchareplacementforGDP.As

DehleyandKroll(2013)haveobserved,thereareessentiallythreewaysthatthiscanbedone:by

“healingtheGDP”,by“complementingtheGDP”,andby“replacingtheGDP.”

Anearlyattemptto“heal”theGDPwascarriedoutbyNordhausandTobin(1973).Theydevelopeda

measureofeconomicwelfare(MEW)thatstartedwithGDPbutaddedinthevalueofleisuretimeand

unpaidworkandsubtractedthevalueofenvironmentaldamage.AnotherattemptathealingtheGDP

wasthegenuineprogressindicator,orGPI,thatwasintroducedbyDalyandCobb(1989).Thisisa

popularreplacementofGDPthat“allowsfortheidentificationofwhichfactorsincreaseordecrease

economicwelfare.”UsingGPIasthemeasureofeconomicwelfare,Kubiszewskietal(2013)wereable

touseestimatesfor17countriesrepresentingoverhalftheworld’spopulationtoshowthatglobal

economicwelfarehasbeenstagnantsince1978.

However,sincewell-beingismulti-dimensional,itmaybemoreappropriatetocomplementGDPwith

otherindicators.PerhapsthemostpopularmeasureofprogressthatcomplementsGDPistheHuman

DevelopmentIndex(http://hdr.undp.org/en/content/human-development-index-hdi)whichcombines

GrossNationalIncomepercapitawithlifeexpectancyatbirthandaneducationindex(meanyearsof

schoolingandexpectedyearsofschooling)

AmorerecentcompositeindexthatismeanttocomplementGDPistheOECD’sBetterLifeIndex

(http://www.oecdbetterlifeindex.org/).Thisindex,launchedbytheOECDin2011asapartofitsBetter

LifeInitiativeandinconjunctionwithits50thAnniversary,consistsof11dimensionsandatotalof24

indicators.Thedimensionscanbegroupedintotwobroadcategories:MaterialLivingConditionsand

QualityofLife.Housing,incomeandjobsarethethreedimensionsincludedunderMaterialLiving

Conditions,whilecommunity,education,environment,civicengagement,health,lifesatisfaction,safety

andwork-lifebalancearetheeightdimensionsincludedunderQualityofLife.Eachdimensionconsists

ofoneormoreindicators,whicharenormalizedandaggregatedinastandardwaytoproduceanoverall

compositeindexofwell-being.Intheirstudyofwell-beingindices,DehleyandKroll(2013)showedthat

theOECD’sBetterLifeIndexwasinfactabetterpredictorofsubjectivewell-beingthatGrossNational

Income.

AnexampleofameasurethattriestoreplacetheGDPaltogetheristheHappyPlanetIndex

(http://www.happyplanetindex.org/).Thisindexcombineslifeexpectancy,experiencedwell-being(as

measuredbythe“LadderofLife”questionontheGallupWorldPoll)andecologicalfootprint(the

amountoflandpercapitaneededtosustainconsumption).

Howdoweknowhowhappypeopleare?Easterlin(1974)usedself-reportedhappinessdatafromtheU.

S.GeneralSocialSurveyinwhichhappinesswasmeasuredbyasinglequestion:

Takenalltogether,howwouldyousaythingsarethesedays—wouldyousaythatyouarevery

happy,prettyhappy,ornottoohappy?

Thiswasabolddeparturefromtraditionaleconomictheorywhichtendedtorelyon“revealed

preferences”todeterminewhatmakespeoplehappy.Economistswereskepticalaboutusing

“subjectivewell-being”dataatfirst(andmanystillare),butinthepastfewdecadesanentirenewfield

of“happinesseconomics”hasemerged.

Canwetrustthemeasuresofhappinessthatareproducedbysurveysofthissort?Afteralllife

satisfactionisasubjectivemeasurethatcannotbeobjectivelyverified.AsKahnemanandKrueger(2006)

statelifesatisfactionis

aglobalretrospectivejudgment,whichinmostcasesisconstructedonlywhenaskedandis

determinedinpartbytherespondent’scurrentmoodandmemory,andbytheimmediate

context.

Itmaybethatreportedlifesatisfactioncanbeinfluencedbysomethingtriviallikethecurrentweather.

Indeed,astudybySchwarzandClore(1983)claimstoshowthatpeoplereportedhigherlifesatisfaction

whentheweatherwassunny.Thesameresultwasfoundinamorerecentstudy(KämpferandMutz,

2013).Lifesatisfactionscoresmayalsobeinfluencedbytheorderofquestionsonasurvey,particularly

whenthelifesatisfactionquestionisprecededbyquestionsthatcanaffectthemoodoftherespondent.

Forexample,intheiranalysisoflifesatisfactionontheCanadianGeneralSocialSurveyfrom2003to

2011,Bonikowskaetal(2013)noticedthatthelowestlevelsoflifesatisfaction(inparticularthe

percentageofrespondentsreportingthehighestscoreof10)occurredin2005and2010.Duringthose

yearsthelifesatisfactionquestionwasprecededbyquestionsontimeuseandtheauthorsspeculate

thatthisemphasisonanegativeaspectofpeople’slivesdeflatedthenumberofresponsesinthe

highestcategories.Inthesamepaper,theauthorsalsonotethatthehighestpercentageofpeople

respondingwithascoreof10occurredin2009whenthelifesatisfactionquestionwasproceededby

questionsoncriminalvictimization.Thespeculationhereisthatthepeoplesurveyedwhowere

remindedofthefactthattheywerenotvictimizedweremorelikelyboostedbythethoughtofbeing

fortunateandratedtheirlifesatisfactionhigherasaresult.Thepercentageofpeoplerespondingwith

10in2009(withvictimizationquestions)was29.7%,whereasthepercentageofpeoplerespondingwith

10in2010(withtimeusequestions)was14.6%.Itishighlyunlikelythatsuchalargedecreasecould

havebeencausedbyanythingotherthancontextualfactors.

However,therehavebeennumerousstudieswhichhavetestedtherobustnessofthesesubjective

measuresoflifesatisfactionsfromsurveys.Forexample,KruegerandSchkade(2008)showthatthereis

enoughconsistencyinlifesatisfactionresponsesthroughtimethatwouldallowtheresultstobeused

forstatisticalanalysis.OswaldandWu,S.(2010)lookedatmeasuresofwell-beingthroughoutthe

UnitedStatesandfoundastrongcorrelationbetweensubjectiveandobjectivemeasuresofwell-being.

Theauthorsconcludethat“subjectivewell-beingdatacontaingenuineinformationaboutthequalityof

humanlives.”

TheDeterminantsofLifeSatisfaction

TheBetterLifeIndexisdefinedasalinearaggregationoftheindividualdimensionsandindicators.But

what should theseweightsbe?Whenweightsareassigned theyact as coefficients that “embody the

relative importance attached to each component” (Foster, McGillivray and Seth, 2013). However,

determiningwhatthoseweightsshouldbeforamultidimensionalindexisaverydifficultproblemthat

can be approached using many different techniques (Decancq and Lugo, 2013). The OECD does not

makeanyattempttodeterminethedimensionweights:itsetsallweightsequaltooneinitiallyandlets

theuserdecideonthe“importance”ofeachdimensionbyadjustingtheindividualslidersfrom0to5.

AsofMarch7,2014usershadsubmittedatotalof66,639 indicesthroughthe interactiveonlinetool.

(Seehttp://blirt.oecdcode.org/).

Figure1Averageuserdefinedweightsforthe11dimensions(Allresponses)

Figure2Averageuserdefinedweightsforthe11dimensions(Canadianresponses)

Nowitshouldbenotedthatthecollecteduserdefinedvectorsofweightsdonotrepresentastatistically

representativesampleandshouldbeusedwithcaution.However,theOECD’sdecisiontoletusers

submittheirownweightssuggeststhattheconstructorsoftheindexthinkthatindividualsarethebest

judgesofwhatmakesthemhappy.AswecanseeinFigures1and2above,thetoptwodimensionsare

0.0679

0.08100.0881 0.0897 0.0898 0.0911 0.0916 0.0947 0.0989

0.1035 0.1037

0.0659

0.08010.0881 0.0889 0.0899 0.0906 0.0931 0.0953 0.0988 0.1030 0.1063

HealthandLifeSatisfaction.TheLifeSatisfactiondimensionconsistsofasingleindicator:theaverage

scorefrom0to10ontheGallupWorldPoll’slifesatisfactionquestion.TheGallupWorldPolluses

Cantril’sLadderandaskspeopletorateoftheircurrentliferelativetothebest(10)andworst(0)

possiblelivestheycanimagine.

ItisinterestingthatLifeSatisfactionisrankedfirstoverallbyusers.HelliwellandPutnam(2004)assert

that“acasecanbemadethattheultimate‘dependentvariable’insocialscienceshouldbehumanwell-

being,andinparticular,well-beingasdefinedbytheindividualherself,or‘subjectivewell-being’.

Perhapswecanuselifesatisfaction(whichhasbeenusedinterchangeablywithsubjectivewell-being

andhappiness)asaproxyforobjectivewell-being.

Ifwefollowthisadviceandtaketheresponsetoalifesatisfactionquestiononasurvey(suchasthe

CanadianCommunityHealthSurvey)asourdependentvariableandregressonaselectionof

demographicandsocio-economicexplanatoryvariables,weshouldbeabletoestimatetherelative

importanceofthevariousdriversofwell-being.TheOECDGuidelinesonMeasuringSubjectiveWell-

being(OECD2013)states:

Ifitisacceptedthatmeasuresofsubjectivewell-beingarevalid,andthattheyaccurately

capturetheconceptsthattheyclaimtomeasure–anoverallevaluationoflifeorthe

experiencedmoodsandemotionsofanindividualoveraperiodoftime–thenitfollowsthat

suchmeasurescanbeusedtoprovideinformationontherelativecontributionofdifferent

factorsandcircumstancestoaperson’swell-being…

UsingtheGallupWorldPoll,forexample,Boarinietal.(2012)didanordinaryleastsquaresregressionof

lifesatisfactionasthedependentvariableagainstalistofdemographicandsocio-economic

independentvariablesthatrepresentedthevariousdomainsoftheOECDBetterLifeIndex(acomplete

listofBetterLifeIndexdimensionsandindicatorsisgiveninTable1).Theyfoundforinstancethatthe

coefficientofthelogbasetwoofhouseholdincomewas0.1482whilethecoefficientofbeingmarried

was0.2584.Fromthiswecanconcludethatbeingmarriedis0.2584/0.1482=1.7435timesas

“important”asthelogofhouseholdincomeindetermininglifesatisfaction.

Table1:DimensionsandIndicatorsfortheOECDBetterLifeIndex

Dimension Indicator

Housing Dwellingswithoutbasicfacilities

Housingexpenditure

Roomsperperson

Income Householdnetadjusteddisposableincome

Householdnetfinancialwealth

Jobs Employmentrate

Jobsecurity

Long-termunemploymentrate

Personalearnings

Community Qualityofsupportnetwork

Education Educationalattainment

Studentskills

Yearsineducation

Environment Airpollution

Waterquality

Civicengagement Consultationonrule-making

Voterturnout

Health Lifeexpectancy

Self-reportedhealth

LifeSatisfaction Lifesatisfaction

Safety Assaultrate

Homiciderate

Work-LifeBalance Employeesworkingverylonghours

Timedevotedtoleisureandpersonalcare

Thedatacomefromthe2011-12CanadianCommunityHealthSurvey(CCHS).TheCCHSisanational,

cross-sectionalsurveyadministeredbyStatisticsCanadathatqueriesthehealthstatus,healthcare

utilizationandotherhealthdeterminantsofCanadians.Itisanationalsurveythatcoversabout98%of

theCanadianpopulation,excludingpersonsonIndianreserves,personsinprisonsandinstitutions,and

alsoexcludesmembersoftheCanadianArmedForces.Inparticular,dataforthisstudycomesfromthe

ShareFileofthe2011-12CCHSandonlycontainsdatafortheprovinceofNewfoundlandandLabrador.

Thesharefilecontainsdataforthoserespondentswhohaveagreedtosharetheirpersonaldatawith

StatisticsCanadapartnersandhadasamplesizeof3,415.Forthepurposesofthisstudythesamplewas

restrictedtorespondentsaged15andover,givingasampleof3,267.

SincetheCCHSisnotasimplerandomsampleandfollowsacomplexsurveydesign,thestrataand

primarysurveyunits(PSUs)mustbetakenintoaccountwhencalculatingstandarderrors.However,to

protecttheconfidentialityofthesurveyparticipantsStatisticsCanadadoesnotidentifythestrataand

PSUsused.Instead,theyprovideasetof500bootstrapweightsthatcanbeusedtocalculatevalid

standarderrorsinstatisticalanalysisandregression.

Measures

Lifesatisfaction

Thedependentvariableusedtomeasurelifesatisfactioncomesfromthequestionaskedofparticipants

onthe2011-12CCHS(variableGEN_02A2):

Usingascaleof0to10,where0means"Verydissatisfied"and10means"Verysatisfied",how

doyoufeelaboutyourlifeasawholerightnow?

ThedistributionofresponsesaredisplayedinFigure3.

Figure3:Frequencyofresponsestothelifesatisfactionquestion,2011-12CCHS,NewfoundlandandLabrador

Priortothe2009-10CCHSsurvey,thislifesatisfactionquestionwasmeasuredona5pointscalewith

respondentsbeingaskedtoratetheirlifesatisfactionfrom“Verydissatisfied”to“Verysatisfied’.To

comparetheresultsofthisquestionovertime,theCCHSincludesaderivedvariable(GENGSWL)that

combinescategoriesfroma0-10toa5-pointscaleasfollows:

• Aresponse0and1becomes“Verydissatisfied”

• Responsesof2,3,and4become“Dissatisfied”

• Aresponseof5isconsidered“Neithersatisfiednordissatisfied”

• Responsesof6,7,and8areconsidered“Satisfied”,and

• Responsesof9and10aregiventopratingof“Verysatisfied”

However,inthefollowingchart(whereweareusingthe5-pointderivedvariablefortheyears2009-

2012)wecanclearlyseethatthepercentageofrespondentswhowere“Verysatisfied”risessharply

from2008to2009,andthatthepercentageofrespondentswhowere“Satisfied”decreasessharply

from2008to2009.Sincethischangeisprobablyduetotheconversionfromthe0-10scaletothe5-

pontscaleandnotduetoanyrealchangeinlifesatisfaction,for2009to2012itisbesttoassumethat

the0-10pointscalegivesthebestmeasureoflifesatisfactionandthat“collapsingofcategories”to

producea5-pointscalemaynotgiveameasureoflifesatisfactionthatistrulycomparablewiththe5-

pointscaleusedinpreviousyears.

Figure3:LifeSatisfaction,CCHS,NewfoundlandandLabrador2003-2012

Thatsaid,forthepurposesofthisstudy,wewillcollapsethecategoriestofacilitatetheuseofordered

logisticandthegeneralizedorderedlogisticregression.FollowingStatisticsCanada’sconversionrulesto

derivea5-pointscalerangingfrom“Verydissatisfied”,“Dissatisfied”,“Neithersatisfiednordissatisfied”,

“Satisfied”and“Verydissatisfied”.However,withthesmallnumberofresponsesinthebottomthree

categories,wehavecombinedtheVerydissatisfied,DissatisfiedandNethersatisfiednordissatisfied

categoriesintoasinglecategorycalled“Notsatisfied”.

2 0 0 3 2 0 0 5 2 0 0 7 2 0 0 8 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2

VerySATISFIED SATISFIED NeitherSAT/DISS DISSATISFIED VeryDISSATIS.

Figure4:Frequencyoflifesatisfaction(5pointscale),CCHS2011-12,NewfoundlandandLabrador

Covariates

Researchonlifesatisfactionandhappinesshasrevealedthattherearemanysocio-demographicfactors

thatdeterminesubjectivewell-being;age,gender,income,maritalstatus,andemploymentstatus,for

example,havebeenshowntoaffectanindividual’swell-being(Clark&Oswald,1994;Layard,2005;

Blanchflower&Oswald,2008;Dolenetal,2008).Allofthesedeterminants,andothers,areincludedas

covariatesinthepresentstudy.

IndependentVariablesandLifeSatisfaction

Weincludegender,age(years),andnumberofpersonsunder12inthehousehold.Forincomewetake

thelogbasetwoofhouseholdincomeanddividebythesquarerootofhouseholdsizetogetthelog

basetwoofhouseholdequivalentincome.Wealsoincludemaritalstatus(married,livingcommon-law,

widowed,separatedordivorced,andsingle),educationalattainment(lessthansecondary,secondary,

somepost-secondary,collegediploma,anduniversitydegree),employmentstatus(employed,

unemployed),andhouseownership(ownerandrenter).

Weincludesomelifestylevariablesaswell.ThesearePhysicalactivity(Inactive,Moderatelyactive,and

Active),smokinganddrinking(wherewehavedefinedaheavydrinkerassomeonewhoconsumesmore

than5drinksatleastonceamonth).

Achronicconditionsvariableisincluded(apersonhasachronicconditioniftheyansweredyestoanyof

thechronicconditionsquestionsontheCCHSsurvey)andthefoodsecurityvariablewithcategories

FoodsecureandModerately/Severelyfoodsecurewasincludedaswell.

Finally,weincludesubjectivemeasuresofwell-being.Theseincludeself-ratedhealthstatus(ordinal

variablewithrankingPoor,Fair,Good,VeryGoodandExcellent),self–ratedmentalhealthstatus(with

categoriesPoor/Fair,Good,VeryGoodandExcellent),self-ratedlifestress(ordinalvariablewith

categoriesNostress,Notverymuchstress,Abitofstress,andQuiteabit/extremeamountofstress),

andsenseofbelongingtocommunity(Veryweak,Somewhatweak,Somewhatstrong,andVerystrong),

SeeTable3belowforabriefdescriptionofthevariablesofinterest.

Table2:DescriptiveSummaryofVariablesofInterest

Obs Mean Std.Dev. Min Max

Gender(reference:Men) 3267 0.442608 0.496771 0 1Women 3267 0.557392 0.496771 0 1

Age 3267 50.49801 18.80555 15 94

Childrenunder12(reference:None) 3267 0.835323 0.370946 0 1One 3267 0.093358 0.290978 0 1Twoormore 3267 0.071319 0.257397 0 1

Logbasetwohouseholdequivalentincome 3263 14.42121 1.07848 3.807355 19.93157

Education(reference:Lessthansecondary) 3191 0.284237 0.451121 0 1Secondary 3191 0.179568 0.383887 0 1Somepostsecondary 3191 0.025384 0.157313 0 1Collegediploma 3191 0.396428 0.489232 0 1Universitydegree 3191 0.114384 0.318327 0 1

Maritalstatus(reference:Married) 3263 0.499234 0.500076 0 1Livingcommon-law 3263 0.080601 0.272263 0 1Widowed 3263 0.107263 0.309495 0 1

Divorcedorseparated 3263 0.074778 0.263073 0 1Single 3263 0.238124 0.426001 0 1

Employed(reference:Employed) 2959 0.515377 0.499848 0 1Unemployed 2959 0.484623 0.499848 0 1

Owner(reference:Renter) 3212 0.16812 0.374031 0 1Owner 3212 0.83188 0.374031 0 1

Physicalactivity(reference:

Inactive) 3178 0.526746 0.499363 0 1Moderatelyactive 3178 0.245437 0.430414 0 1Active 3178 0.227816 0.419489 0 1

Smoking(reference:Non-smoker) 3263 0.775667 0.417207 0 1Smoker 3263 0.224333 0.417207 0 1

Heavydrinking(reference:Notheavydrinker) 3234 0.768089 0.422118 0 1Heavydrinker 3234 0.231911 0.422118 0 1

Chronicconditions(reference:Nochroniccondition) 3266 0.36038 0.480184 0 1Chroniccondition 3266 0.63962 0.480184 0 1

Foodsecurity(reference:Foodsecure) 3201 0.937832 0.241498 0 1Moderately/Severelyfoodinsecure 3201 0.062168 0.241498 0 1

Self-ratedhealth(reference:poor) 3263 0.037695 0.190488 0 1Fair 3263 0.102666 0.303569 0 1Good 3263 0.261416 0.439473 0 1Verygood 3263 0.439473 0.496399 0 1Excellent 3263 0.15875 0.365498 0 1

Self-ratedmentalhealth(reference:Poor/Fair) 3173 0.041916 0.200429 0 1Good 3173 0.221872 0.415571 0 1VeryGood 3173 0.363694 0.481138 0 1Excellent 3173 0.372518 0.483552 0 1

Senseofbelongingtocommunity(reference:Veryweak) 3247 0.161996 0.368504 0 1Somewhatweak 3247 0.282107 0.450094 0 1Somewhatstrong 3247 0.431167 0.495316 0 1Verystrong 3247 0.124731 0.330464 0 1

Self-ratedlifestress(reference:Nostress) 3146 0.045137 0.207637 0 1Notverymuch 3146 0.146217 0.35338 0 1Abitofstress 3146 0.535919 0.498788 0 1Quiteabit/extreme 3146 0.272727 0.445433 0 1

Regression

Wefirstapplyordinaryleastsquarestodetermineeffectsofthecovariatesonlifesatisfaction.Although

lifesatisfactionisanordinalvariablewhichwouldseemtorequireanordinalregressionmethod,the

CCHSquestionthatmeasureslifesatisfactionhas11responsecategoriesonanumericalscalefrom0to

10.Withthismanylevelsofthedependentvariable,ordinalregressionwouldbeverycumbersome

(unlessthecategorieswerecollapsed),soweemployordinaryleastsquaresasafirststepinmodelling

theeffectsofthepredictingvariablesonlifesatisfaction.Ordinaryleastsquaresregressionrequiresthat

theindependentvariableiscardinal.Thiswouldimply,forinstance,thatapersonwhorespondedwith

an8onthelifesatisfactionquestionis“twice”ashappyassomeonewhorespondedwitha4.Itis

impossibletoverifythisassumption;neverthelessitisoftenmadeandordinaryleastsquaresregression

isoftenapplied.

Thestandardregressiontechniquestousewouldbeorderedlogisticorprobitregression.Sharpeetal

(2010),forexample,usedorderedprobitregressiontodoanextensiveanalysisoflifesatisfactionof

CanadiansusingtheCanadianCommunityHealthSurveyof2007-08.Fortheorderedlogisticregression

model,ourdependentvariablewillbelifesatisfactionwiththecategoriesfrom0to11collapsedinto

threecategories(0–5collapsedtobecomethecategory“Notsatisfied”,6-8collapsedtobecomethe

category“Satisfied”,and9–10collapsedtobecome“Verysatisfied”).

Oneofthekeyassumptionsoftheorderedlogisticregressionmodelisthattheparametersdonotdiffer

acrossthecategoriesofthedependentvariable.Thisisoftenreferredtoastheparallellines

assumption.Forthecumulativeorderedlogisticregressionmodel,thismeansthatforanordinal

dependentvariableYwithMcategoriesthemodelcanbewrittenas:

𝑃 𝑌# > 𝑗 = 𝑔 𝑋𝛽 =exp(𝛼0 + 𝑋#𝛽)

1 + [exp 𝛼0 + 𝑋#𝛽 ], 𝑗 = 1,2, … ,𝑀 − 1

Notethatthebetacoefficientsarethesameforalllevelsofthedependentvariable(for𝑗 =

1,2, … ,𝑀 − 1).

Thisassumptionisoftenviolated,andwhenitis,theresultsoftheorderedlogisticregression

modelmaybeincorrect(Williams,2006)anditmaybemoreappropriatetousethegeneralized

orderedlogisticregressionmodel.Thiscanbewrittenas:

𝑃 𝑌# > 𝑗 = 𝑔 𝑋𝛽0 =exp(𝛼0 + 𝑋#𝛽0)

1 + [exp 𝛼0 + 𝑋#𝛽0 ], 𝑗 = 1,2, … ,𝑀 − 1

Notethateachleveljofthedependentvariablehasitsownsetofbetacoefficients.

TheprobabilitiesthatYwillbeequaltothevalues𝑗 = 1,2, … ,𝑀 − 1aregivenby:

𝑃 𝑌# = 1 = 1 − 𝑔 𝑋𝛽;

𝑃 𝑌# = 𝑗 = 𝑔 𝑋#𝛽0<; − 𝑔 𝑋#𝛽0 , 𝑗 = 1,2, … ,𝑀 − 1

𝑃 𝑌# = 𝑀 = 𝑔 𝑋#𝛽=<;

ThismodelcanbeestimatedinSTATAbytheuser-writtengologit2command(Williams,2006)

whichwillcomputeadifferentsetparametersforeachlevelofthedependentvariable.The

autofitoptionofthegologit2commandwillperformaseriesofWaldteststoseeifanyofthe

variablesviolatetheparallellinesassumption.Thiswillproduceamoreparsimoniousmodel

sinceitwillcomputedifferentsetsofcoefficientsforeachlevelofthedependentvariableonly

forthoseindependentvariablesthatfailtomeettheparallellinesassumption.

Thegologit2programhasbeenusedtostudythedeterminantsoflifesatisfactionbyVinsonand

Ericson(2014),Liang, Yamashita, and Brown (2013) and AmbreyandFleming(2013).

Aswedidfororderedlogisticregression,forthegeneralizedorderedlogisticregressionweuse

thecollapsedlifesatisfactionvariablewiththreelevels(“Notsatisfied”,“Satisfied”,and“Very

satisfied”)asthedependentvariable.Thegologit2command,however,doesnotworkwith

bootstrapweightssowehadtousegologit3,whichisabetaprogramwrittenbyRichard

Williamsthatdoesworkwithbootstrapweightsandproducesthesameresultsasgologit2.The

gologit3commandrequiresthegolsvysubroutineinordertoworkwiththecomplexsurvey

designoftheCCHS.

BivariateRegressionResults

Wefirstappliedordinaryleastsquares,orderedlogisticandgeneralizedorderedlogistics

regressiononthedependentvariable,lifesatisfaction,andeachoftheindependentvariables

separately.Thefollowingtableshowtheresultsoftheindividualpairwiseregressions(we

reportoddsratiosfortheorderedandgeneralizedorderedlogisticregression).

Table4:BivariateRegressionResults

OrderedLogisticRegression(OR) GeneralizedOrderedLogisticsRegression(OR)

Notsatisfiedvs(SatisfiedandVerysatisfied)

(NotsatisfiedandSatisfied)vsVerysatisfied

Gender(reference:Men) Women 0.0492 1.060 0.861 1.100

Age 0.00172 1.004 0.983*** 1.007***

Childrenunder12(reference:None)

One 0.228** 1.315* 2.902** 1.211Twoormore 0.221 1.499** 1.812 1.461**

Logbasetwohouseholdequivalentincome 0.171*** 1.207*** 1.443*** 1.154***

Education(reference:Lessthansecondary)

Secondary -0.0506 0.818 1.526 0.719**Somepostsecondary 0.0123 0.867 2.140 0.732Collegediploma 0.0733 1.018 1.655** 0.925Universitydegree 0.155 1.264 2.052* 1.154

Maritalstatus(reference:Married)

Livingcommon-law -0.205 0.806 1.372 0.750Widowed -0.385** 0.689* 0.460*** 0.778

Divorcedorseparated -0.865*** 0.326*** 0.255*** 0.419***Single -0.476*** 0.577*** 1.010 0.512***

Employed(reference:Employed)

Unemployed -0.239*** 0.828* 0.391*** 0.943

Owner(reference:Renter) Owner 0.451*** 1.837*** 1.582** 1.922***

Physicalactivity(reference:Inactive)

Moderatelyactive 0.191** 1.181 2.124*** 1.061Active 0.446*** 1.538*** 2.936*** 1.401***

Smoking(reference:Non-smoker)

Smoker -0.483*** 0.645*** 0.405*** 0.726**

Heavydrinking(reference:Notheavydrinker)

Heavydrinker -0.155* 0.779** 1.011 0.739**

Chronicconditions(reference:Nochroniccondition)

Chroniccondition -0.299*** 0.792** 0.287*** 0.908

Foodsecurity(reference:Foodsecure)

Moderately/Severelyfoodinsecure -1.068*** 0.260*** 0.223*** 0.302***

Self-ratedhealth(reference:poor)

Fair 1.573*** 3.341*** 3.014*** 1.373Good 2.267*** 8.393*** 12.88*** 2.109**Verygood 2.741*** 15.37*** 34.76*** 3.811***Excellent 3.247*** 29.64*** 65.05*** 7.538***

Self-ratedmentalhealth(reference:Poor/Fair)

Good 1.508*** 5.822*** 5.387*** 2.999**VeryGood 1.825*** 8.199*** 13.86*** 3.614**Excellent 2.574*** 23.16*** 30.63*** 10.81***

Senseofbelongingtocommunity(reference:Veryweak)

Somewhatweak 0.332 1.337 1.728 1.107Somewhatstrong 0.811*** 2.317*** 3.877*** 1.803**

Verystrong 1.153*** 3.861*** 5.477*** 3.079***

Self-ratedlifestress(reference:Nostress)

Notverymuch -0.439*** 0.447*** 0.766 0.428***Abitofstress -0.901*** 0.253*** 0.283*** 0.247***Quiteabit/extreme -1.696*** 0.123*** 0.0817*** 0.170***

Itisperhapssurprisingtoseethateducationdoesnothaveasignificanteffectonlifesatisfaction;in

fact,aswecanseeinFigure5below,respondentswithsomepost-secondaryeducationreporteda

slightlylowermeanlifesatisfactionthanrespondentswhohadlessthanahighschooleducation.

Figure5:Educationandlifesatisfaction

Themostsignificantvariablesinthepairwiseregressionswouldseemtomaritalstatusandthe

subjectivevariablesself-ratedhealth,self-ratedmentalhealth,senseofbelongingto

communityandself-ratedlifestress.

Figure6:Maritalstatusandlifesatisfaction

8 8.17.9

8.3 8.3

Less than high school High school Some postsecondary College diploma University degree

Formaritalstatuswenoteinparticularthatdivorcedorseparatedindividualshaveameanlife

satisfactionscoreof7.4comparedtoameanscoreof8.4forthosewhoaremarriedorliving

common-law.

Figure7:Self-ratedhealthandlifesatisfaction

8.4 8.47.9

Life s

atisfa

Married Living common-law Widowed Divorced or separated Single

Poor Fair Good Very good Excellent

Self-ratedhealthishighlysignificant.TheoddsratioforrespondentswithExcellentself-rated

healthis29.64comparedtothebasecategoryofrespondentswithPoorself-ratedhealth.

Householdincomeissignificantandwenotethattheoddsratiosinthegeneralizedordered

logisticregressionsforincomeare1.443forNotsatisfiedvs(SatisfiedandVerysatisfied)and

1.154for(NotsatisfiedandSatisfied)vsVerysatisfied.Howdoweinterpretthesecoefficients?

Thesecoefficientsindicatethattheeffectofincomeonlifesatisfactionsdependsonwherethe

respondentisinthelifesatisfactionoutcomedistribution.Adoublingofincomewouldmean

moretosomeoneonthelowerendofthelifesatisfactionscalethantosomeoneontheupper

end.TheoddsofbeingSatisfiedorVerysatisfiedasopposedtoNotsatisfiedis1.443whilethe

oddsofbeingVerysatisfiedasopposedtoNotsatisfiedorSatisfiedisonly1.154.AsBoesand

Winkelmann(2004)ask“isitpossiblethattheeffectofincomeonhappinessisdifferentin

differentpartsoftheoutcomedistribution?Coulditbethat“moneycannotbuyhappiness,but

buy-offunhappiness”asaproverbsays?”AlthoughBoesandWinkelmanndonotusethe

gologit2procedure,wecanseeclearlyitcancertainlybeusedtoanswertheirquestion.

MultivariateRegressionResults

Weproducefourdifferentmodels.InModel1weincludejustthestandardsocio-demographic

variablesandalsoincludethevariableage2by100whichisthesquareoftheagedividedby

100.InModel2weaddinthelifestylevariablesphysicalactivity,smokingandheavydrinking.

InModel3weaddinchronicconditionsandfoodsecurityvariable.Finally,inModel4weadd

inallthesubjectivevariables:self-ratedhealth,self-ratedmentalhealth,sel-ratedlifestress

andsenseofbelongingtocommunity.

Table5:Model1

OrderedLogisticRegression GeneralizedOrderedLogisticsRegression

Gender(reference:Men)

Women 0.0798 1.105 0.902 1.130

Age -0.0636*** 0.942*** 0.869*** 0.961*

age2by100 0.0698*** 1.073*** 1.164*** 1.050*

One 0.282** 1.486** 3.225** 1.345Twoormore 0.308* 1.755*** 2.101 1.665**

Logbasetwohouseholdequivalentincome 0.128*** 1.168** 1.392*** 1.114*

Secondary -0.123 0.726** 1.175 0.637***Somepostsecondary 0.0440 0.948 1.629 0.832Collegediploma -0.109 0.838 1.199 0.763*Universitydegree -0.150 0.903 1.074 0.878

Livingcommon-law -0.215 0.886 1.062 0.851Widowed -0.420* 0.666 0.579 0.733Divorcedorseparated -0.763*** 0.356*** 0.279*** 0.471***

Single -0.567*** 0.624*** 0.784 0.576***

Unemployed -0.263*** 0.794* 0.431*** 0.901

Owner(reference:Renter)

Owner 0.251** 1.478** 1.285 1.554**

Table6:Model2

OrderedLogisticRegression

GeneralizedOrderedLogisticsRegression

Women 0.0423 1.035 0.830 1.062

Age -0.0434*** 0.960* 0.910** 0.974

age2by100 0.0484*** 1.051** 1.107** 1.034

One 0.317*** 1.522** 3.113** 1.365Twoormore 0.294* 1.717*** 1.864 1.628**

Logbasetwohouseholdequivalentincome 0.0985** 1.135** 1.302** 1.096

Secondary -0.13 0.733* 1.099 0.645***Somepostsecondary 0.0217 0.95 1.679 0.840Collegediploma -0.109 0.849 1.191 0.769*Universitydegree -0.25 0.836 0.884 0.835

Livingcommon-law -0.162 0.906 1.386 0.840Widowed -0.427* 0.653 0.626 0.706

Divorcedorseparated -0.779*** 0.343*** 0.302*** 0.449***Single -0.542*** 0.617** 0.865 0.559***

Unemployed -0.281*** 0.769** 0.410*** 0.876

Owner 0.181 1.396* 1.211 1.481**

Moderatelyactive 0.152 1.152 2.016** 1.035Active 0.513*** 1.733*** 2.437*** 1.621***

Smoker -0.277** 0.838 0.447*** 0.981

Heavydrinker -0.155 0.764** 0.822 0.740**

Table7:Model2

OrderedLogistic

RegressionGeneralizedOrderedLogisticsRegression

Women 0.0548 1.041 0.874 1.066

Age -0.0374** 0.966 0.933 0.978

age2by100 0.0438** 1.046* 1.083* 1.03

One 0.255** 1.414* 2.712** 1.279Twoormore 0.182 1.510* 1.494 1.477*

Logbasetwohouseholdequivalentincome 0.0378 1.058 1.169 1.035

Secondary -0.117 0.746* 1.056 0.651***Somepostsecondary 0.0208 0.926 1.678 0.827Collegediploma -0.0851 0.873 1.263 0.782Universitydegree -0.227 0.867 0.847 0.860

Livingcommon-law -0.174 0.892 1.350 0.827Widowed -0.452** 0.635 0.574 0.693

Divorcedorseparated -0.720*** 0.363*** 0.296*** 0.469***Single -0.578*** 0.589*** 0.780 0.535***

Unemployed -0.282*** 0.760** 0.395*** 0.873

Owner 0.0508 1.185 0.834 1.311

Physicalactivity

(reference:Inactive) Moderatelyactive 0.108 1.105 1.790* 1.001

Active 0.474*** 1.687*** 2.222** 1.585***

Smoker -0.239** 0.859 0.462*** 1.005

Heavydrinker -0.188* 0.736** 0.735 0.717**

Chroniccondition -0.327*** 0.737*** 0.401** 0.807*

Moderately/Severelyfoodinsecure -0.794*** 0.359*** 0.254*** 0.423***

Table8:Model4

OrderedLogisticRegression GeneralizedOrderedLogisticRegression

Gender(reference:Men) Women 0.042 1.052 0.760 1.118

Age -0.0253* 0.973 0.900** 0.991

age2by100 0.0242 1.026 1.106* 1.006

One 0.106 1.211 2.609* 1.082

Twoormore 0.0428 1.39 1.615 1.302

Logbasetwohouseholdequivalentincome 0.00921 1.046 1.223 0.995

Secondary -0.171 0.678** 0.918 0.579***Somepostsecondary 0.0981 1.106 2.287 0.825Collegediploma -0.139 0.802 1.311 0.705**Universitydegree -0.261* 0.774 0.952 0.724

Livingcommon-law -0.166 0.904 2.112 0.782Widowed -0.512*** 0.561* 0.501 0.622

Divorcedorseparated -0.604*** 0.323*** 0.362* 0.346***Single -0.552*** 0.531*** 0.905 0.450***

Unemployed -0.0984 0.888 0.638 0.958

Owner -0.042 1.017 0.859 1.136

Moderatelyactive 0.0254 1.063 1.577 0.963Active 0.131 1.166 1.170 1.126

Smoker -0.138 0.914 0.496** 1.090

Heavydrinker -0.216*** 0.678*** 0.533* 0.688**

Chroniccondition -0.0258 1.069 0.762 1.148

Moderately/Severelyfoodinsecure -0.444** 0.514** 0.360*** 0.562*

Self-ratedhealth(reference:Poor)

Fair 1.629*** 3.498** 1.946 1.855Good 1.991*** 6.735*** 6.327*** 2.105Verygood 2.233*** 10.22*** 12.13*** 3.383**Excellent 2.548*** 16.07*** 19.82*** 5.712***

Self-ratedmentalhealth(reference:Poor/Fair)

Good 0.857*** 2.906*** 3.071** 1.837VeryGood 0.882*** 3.056*** 4.570*** 1.774Excellent 1.458*** 7.872*** 10.39*** 4.838**

Senseofbelongingtocommunity(reference:Veryweak)

Somewhatweak 0.0975 1.169 1.767 0.953Somewhatstrong 0.430** 1.913** 4.254*** 1.489Verystrong 0.554*** 2.584*** 5.470*** 1.985**

Self-ratedlifestress(reference:Nostress)

Notverymuch -0.293*** 0.468*** 1.036 0.474***Abitofstress -0.577*** 0.293*** 0.393* 0.293***Quiteabit/extreme -0.953*** 0.209*** 0.140*** 0.288***

Significancelevels:*p<0.10,**p<0.05,***p<0.01

Discussion

Genderisnotsignificantinanyofthemodels.Thecoefficientsforgenderintheordinaryleastsquares

regressionsinModels1,2,3,and4are0.0798,0.0423,0.0548,and0.042,respectively,withnoneofthe

coefficientsbeingstatisticallysignificant.Thisindicatesthatthereisnodifferenceinlifesatisfaction

betweenmalesandfemalesinourdata.Similarly,noneofthecoefficientsforgenderintheordered

logisticandgeneralizedorderedlogisticregressionmodelsaresignificant.

Ageissignificantforlifesatisfaction.InModel1,forinstance,theordinaryleastsquaresregression

coefficientsforageandagesquareddividedby100are-0.0636and0.0698,respectively,withboth

coefficientssignificantatthep<0.01level.Thefactthatthecoefficientofageisnegativeandthe

coefficientofagesquaredispositiveindicatesthatlifesatisfactionis“U-shaped”fortheagevariable,

withlifesatisfactiondroppingasageincreases,reachingaminimumatacertainpointinmiddleageand

thenincreasingonceagainasageincreases.ThecoefficientforageinModel2is-0.0434(significance

levelp<0.01),whilethecoefficientinModel3is-0.0374(significancelevelp<0.05)andforModel4

thecoefficientforageis-0.0253(significancelevelp<0.1).

Theordinaryleastsquarescoefficientforhouseholdequivalentincome(logarithmbasetwo)inModel1

is0.128(significancelevelp<0.01).Thisindicatesthataoneunitincreaseintheincomevariable(which

meansthatthehouseholdincomedoubles)correspondstoa0.128pointincreaseinlifesatisfactionon

the0-10scale.However,asothervariablesareadded,incomebecomeslesssignificant.Infact,in

Models3and4householdincomeisnotsignificantintheordinaryleastsquares,orderedlogisticor

generalizedorderedlogisticregressions.

BeingemployedissignificantforlifesatisfactioninModels1,2and3,but,aswithincome,being

employedceasestobeasignificantfactorwhenthesubjectivevariablesonself-ratedhealth,self-rated

mentalhealth,senseofbelongingtocommunityandself-ratedlifestressareaddedinModel4.Wesee

thesamepatternwiththenumberofchildrenundertwelveandownervariables;havingchildrenunder

twelveandbeingahomeownerissignificantModel1butneitherissignificantinModel4.

Thisisnottrueofmaritalstatus,however,whichissignificantinallfourmodels.Theorderedlogistic

regressionoddsratioforseparatedanddivorcedinModel1is0.356(p<0.01).Thismeansthatfor

someonewhoisseparatedordivorcedtheoddsofbeinginahighercategoryoflifesatisfactionis0.356

timesthatofsomeonewhoismarried.InModel4theorderedlogisticoddsratioforseparatedand

divorcedisevensmallerat0.323(p<0.01).Ifwelookatthegeneralizedorderedlogisticregression,the

oddsratiosforseparatedanddivorcedinModel1are0.279(p<0.01)forNotsatisfiedvsSatisfiedand

Verysatisfiedand0.471(p<0.01)forNotsatisfiedandSatisfied)vsVerysatisfied.Aswithincome,

maritalstatusaffectsthelowerpartofthelifesatisfactiondistributionmorethantheupperpart.The

oddsthatapersonwhoisdivorcedorseparatedwillbeSatisfiedorVerysatisfiedasopposedtoNot

satisfiedisonly0.279timesthatofapersonwhoismarried.Theoddsthatapersonwhoisdivorcedor

separatedwillbeVerysatisfiedasopposedtoNotsatisfiedorsatisfiedareslightlyhigher,at0.471times

thatofapersonwhoismarried.ItisinterestingtonotethoughthatinModel4,whilestillsignificant,

thecoefficientsfordivorcedorseparatedare0.362(p<0.1)forNotsatisfiedvsSatisfiedandVery

satisfiedand0.346(p<0.01)forNotsatisfiedandSatisfiedvsVerysatisfied.Inthiscaseitwouldappear

thatseparationordivorcehasthesameeffectonlifesatisfactionregardlessofwhataperson’slife

satisfactionlevelis.

Ingeneralthen,otherthanformaritalstatus,whenweintroducesubjectivewell-beingvariablesin

Model4,othervariableswhichweresignificantinearliermodelsbecomelessstatisticallysignificantor

notsignificantatall.Beingphysicallyactiveasopposedtobeinginactive,forexample,issignificantin

Models2and3,withtheorderedlogisticregressionoddsratiosof1.733(p<0.01)inModel2and1.687

(p<0.01).However,inModel4,beingphysicallyactiveceasestobesignificantwithanoddsratioof

1.116.AchronichealthconditionvariableisintroducedinModel3andhavingachronicconditionis

significantwithanorderedlogisticregressionoddsratioof0.737(p<0.01).Butwhenweintroduceself-

ratedhealthinModel4,theoddsratioforhavingachronichealthconditiondropsto1.069andisnot

statisticallysignificant.Theself-ratedhealthvariableisverysignificant,withevenfairhealthhavingan

oddsratioof3.498(p<0.05)comparedtothebasecategoryofpoorhealth.Forthosewithexcellent

health,theoddsofbeinginahigherlifesatisfactioncategoryis16.07(p<0.01)timesthatofsomeone

withpoorhealth.Itappearsthatsubjectivefeelingsabouthealthmaybemoreimportantindetermining

lifesatisfactionthanobjectivemeasuresofhealth.Thegeneralizedorderedlogisticregressionodds

ratiosforExcellenthealthinModel4are19.82(p<0.01)forNotsatisfiedvsSatisfiedandVerysatisfied

and5.712(p<0.01)forNotsatisfiedandSatisfiedvsVerysatisfied.Soself-ratedhealthwouldappearto

makeamuchgreaterdifferenceatthelowerendofthelifesatisfactiondistributionthanitdoesatthe

upperend.

Theothersubjectivewell-beingvariable,self-ratedmentalhealth,senseofbelongingtocommunity,and

self-ratedstressarealsohighlysignificant,and,notsurprisingly,thegeneralizedorderedlogistic

regressionoddsratiosindicatethatthesesubjectivevariableshaveagreatereffectonthelowerendof

thelifesatisfactiondistribution.

Finally,wenotethatheavydrinking,definedashavingmorethan5drinksatatimeatleastoncea

month,doesnotbecomestatisticallyinsignificantwhensubjectivewell-beingvariablesareincludedin

themodel.HeavydrinkingissignificantinModel2withorderedlogisticregressionoddsratioof0.764(p

<0.05)andinModel3ithasanoddsratioof0.736(p<0.05).However,inModel4heavydrinking

becomesmoresignificantwithanorderedlogisticregressionoddsratioof0.678(p<0.01).The

generalizedorderedlogisticregressionoddsratiosforheavydrinkingInModel4are0.533(p<0.10)for

NotsatisfiedvsSatisfiedandVerysatisfiedand0.688(p<0.05)forNotsatisfiedandSatisfiedvsVery

satisfied.Inthiscasethereisverylittledifferencebetweentheeffectofheavydrinkingonthelowerend

ofthelifesatisfactiondistributionandontheupperend.

Conclusion

WecanconcludethatthemostimportantdeterminantsoflifesatisfactionCanadianCommunityHealth

SurveyforNewfoundlandandLabradorseemtobesubjectivemeasuresofwell-being.Otherthan

maritalstatus,whichremainshighlysignificantinallfourmodels,mostofthevariablesinModel1are

notsignificantwhensubjectivewell-beingvariablesareintroducedinModel4.Otherthanthe

interpretationofthecoefficients,theresultsfromordinaryleastsquaresandorderedlogisticregression

areverysimilar,andonecouldargue,asothershave,thatthereisnoadvantagetousingordered

logisticregression,particularlywhenthereare11categoriesofthedependentvariable.However,

generalizedorderedlogisticregressionisanimprovementsinceitprovidesdifferentsetofcoefficients

fordifferentlevelsofthelifesatisfactiondistribution.Thisgivesamorecompletepictureofhowthe

explanatoryvariablesaffectthedependentvariable.

Inaddition,ifindependentvariablesaffectdifferentpartsofthelifesatisfactiondistributiondifferently,

thenitmaybeimportanttolookatwaysofmeasuringinequalityinthedistributionoflifesatisfactionin

muchthesamewayaswelookatmeasuringinequalityinthedistributionofincome.Isitmore

importanttoimprovethelifesatisfactionofsomeonewhoiscurrentlydissatisfied,orisitmore

importanttoimprovethelifesatisfactionofsomeonewhoisalreadysatisfiedwithlife?Ifacertain

numberofpeoplewhocurrentlyputtheirlifesatisfactionscoreat8movesuptoascoreof9,thenthe

overallmeanlifesatisfactionscorehasimproved.Butperhapsitismoreimportanttoraisetheoverall

meanlifesatisfactionscorebyimprovingthelifesatisfactionscoreofpeoplewhohaveamuchlower

score.

References

Ambrey,C.,&Fleming,C.(2013).LifesatisfactioninAustralia:EvidencefromtenyearsoftheHILDAsurvey.SocialIndicatorsResearch,.doi:10.1007/s11205-11012-10228-11200.

Boarini,R.,M.Comola,C.Smith,R.ManchinandF.DeKeulenaer(2012).Whatmakesforabetterlife?Thedeterminantsofsubjectivewell-beinginOECDcountries,evidencefromtheGallupWorldPoll.OECDStatisticsWorkingPapers.OECDPublishing.http://dx.doi.org/10.1787/5k9b9ltjm937-en

Boes,S.&Winkelmann,R.(2004).Incomeandhappiness:Newresultsfromgeneralizedthresholdandsequentialmodels.UniversityofZurich,SocioeconomicInstituteworkingpaperno.0407

Boes,S.&Winkelmann,R.(2006a).Theeffectofincomeonpositiveandnegativesubjectivewellbeing.Workingpaper.Availablefrom:http://www.sts.uzh.ch/research/workingpapers/swbinc.pdf

Bonikowska,A.,Helliwell,J.,Hou,F.,&Schellenberg,G.(2013).AnassessmentoflifesatisfactionresponsesonrecentStatisticsCanadasurveys.Analyticalstudiesbranchresearchpaperseries.Ottawa:StatisticsCanada.

Bonikowska,A.,Helliwell,J.,Hou,F.,&Schellenberg,G.2013.AnassessmentoflifesatisfactionresponsesonrecentStatisticsCanadaSurveys.AnalyticalStudiesBranchResearchPaperSeries351:1–38.

Daly,H.E.,CobbJr.,J.B.,1989.FortheCommonGood:RedirectingtheEconomyTowardCommunity,theEnvironment,andaSustainableFuture.BeaconPress,Boston

Decancq,K.,&Lugo,M.A.(2013).Weightsinmultidimensionalindicesofwellbeing:Anoverview.EconometricReviews,32(1),7-34.

Delhey,J.,&Kroll,C.(2012).A‘‘happinesstest’’forthenewmeasuresofnationalwell-being:HowmuchbetterthanGDParethey?.SocialScienceResearchCenterBerlin(WZB):WZBDiscussionPaper

Diener,E.(2006),“GuidelinesforNationalIndicatorsofSubjectiveWell-BeingandIll-Being”,AppliedResearchinQualityofLife,No.1,pp.151-157.

Dolan,P.,Peasgood,T.andWhite,M.(2008),‘Dowereallyknowwhatmakesushappy?Areviewoftheeconomicliteratureonthefactorsassociatedwithsubjectivewell-being’,JournalofEconomicPsychology,29:1,94–122.

Easterlin(1974).DoesEconomicGrowthImprovetheHumanLot?SomeEmpiricalEvidence.InPaulA.DavidandMelvinW.Reder,eds.,NationsandHouseholdsinEconomicGrowth:EssaysinHonorofMosesAbramovitz,NewYork:AcademicPress,Inc.

Foster,J.,McGillivray,M.&Seth,S.(2013)CompositeIndices:RankRobustness,StatisticalAssociation,andRedundancy.EconometricReviews,32:1,35-56,DOI:10.1080/07474938.2012.690647

Helliwell,J.F.(2008).Lifesatisfactionandqualityofdevelopment(No.w14507).NationalBureauofEconomicResearch.

Kahneman,D.,&Krueger,A.B.(2006).Developmentsinthemeasurementofsubjectivewell-being.JEconPersp20(1):3–24

Kämpfer,S.,&Mutz,M.(2013).Onthesunnysideoflife:Sunshineeffectsonlifesatisfaction.Socialindicatorsresearch,110(2),579-595.Krueger,A.B.,&Schkade,D.A.(2008)‘Thereliabilityofsubjectivewellbeingmeasures.’JournalofPublicEconomics92(8-9),1833–18454Kubiszewski,I.,Costanza,R.,Franco,C.,Lawn,P.,Talberth,J.,Jackson,T.,&Aylmer,C.(2013).BeyondGDP:Measuringandachievingglobalgenuineprogress.EcologicalEconomics,93,57-68.

Layard,R.(2005),Happiness:LessonsfromaNewScience,London:AllenLane.

Liang,J.,Yamashita,T.,&Brown,J.S.(2013).LeisureSatisfactionandQualityofLifeinChina,Japan,andSouthKorea:AComparativeStudyUsingAsiaBarometer2006.JournalofHappinessStudies,14(3),753-769.

Nordhaus,W.andTobin,J.(1973)“IsGrowthObsolete?”inTheMeasurementofEconomicandSocialPerformance,NationalBureauofEconomicResearch,1973

OECD(2013),OECDGuidelinesonMeasuringSubjectiveWell-being,OECDPublishing.http://dx.doi.org/10.1787/9789264191655-en

Oswald,A.J.,&Wu,S.(2010).Objectiveconfirmationofsubjectivemeasuresofhumanwell-being:EvidencefromtheUSA.Science,327(5965),576-579.

Schwarz,N.,&Clore,G.L.(1983).Mood,misattribution,andjudgmentsofwell-being:Informativeanddirectivefunctionsofaffectivestates.Journalofpersonalityandsocialpsychology,45(3),513.

Sharpe,A.,Ghanghro,A.,Johnson,E.andKidwai,A.(2010),“Doesmoneymatter?DeterminingthehappinessofCanadians”,CSLSResearchReportNo.2010-09,CentrefortheStudyofLivingStandards,Ottawa

Vinson,T.,&Ericson,M.(2014).ThesocialdimensionsofhappinessandlifesatisfactionofAustralians:EvidencefromtheWorldValuesSurvey.InternationalJournalofSocialWelfare,23:240–253.doi:10.1111/ijsw.12062

the determinants of life satisfaction march 2016quality of life. housing, income and jobs are the...

Documents

some new evidence on the determinants of satisfaction

determinants of customer satisfaction and customer loyalty

h201106 - determinants of job...

determinants of customer satisfaction in

determinants of customer satisfaction with the season

determinants of job satisfaction and dissatisfaction among

determinants of customer satisfaction of safaricom mobile

determinants of customers' satisfaction for stock broking...

determinants of residential satisfaction

determinants of consumer satisfaction of mobile commerce

the determinants of customer satisfaction, loyalty and

determinants of job satisfaction and turnover among...

determinants of tourist satisfaction - cultural tourism

cross-country determinants of life satisfaction: exploring...

cross-cultural determinants of job satisfaction

determinants of customer satisfaction in supermarkets in

1992-8645 key determinants of customer satisfaction ... ·...

patients’ satisfaction and its determinants in outpatient...

determinants of customer satisfaction in cellular industry...

determinants of job satisfaction among physicians in