which employers sponsor defined contribution plans ...which employers sponsor defined contribution...
Post on 27-Jun-2020
3 Views
Preview:
TRANSCRIPT
PRELIMINARYANDINCOMPLETEPLEASEDONOTCITEWITHOUTAUTHORPERMISSION
Whichemployerssponsordefinedcontributionplans?
Evidencefromlinkedemployer-employeeadministrativerecords*
MichaelGideon,Ph.D.U.S.CensusBureau
JoshuaMitchell,Ph.D.U.S.CensusBureau
November7,2016
________________________________________
* Thispaperisreleasedtoinforminterestedpartiesofongoingresearchandtoencouragediscussionofworkinprogress.TheanalysisandconclusionssetfortharethoseoftheauthorsanddonotindicateconcurrencewithothermembersoftheresearchstaffoftheU.S.CensusBureau,nordoesitrepresentanyofficialpositionoftheU.S.CensusBureau..Anyerrorsareourown.
ABSTRACTThis paper explores an innovative approach to measuring defined contribution (DC) retirement plan sponsorship using tax administrative records. Using this measure, we estimate lower employer sponsorship rates yet higher worker coverage than estimates from establishment and household surveys. We examine characteristics of this measure by linking it to employer-level data from the Census Bureau’s Business Register, plan-level data from Form 5500s filed annually with the Internal Revenue Service (IRS), and household-level data from the Current Population Survey Annual Social and Economic Supplement (CPS ASEC) and the American Community Survey (ACS).
Using universe administrative records we find that in 2012, 79 percent of all wage and salary employees work for an employer who sponsors a defined contribution (D.C.) plan. Of these workers, we observe that 41 percent made a contribution to a D.C. plan. This works out to an overall measured participation rate in D.C. plans of 32 percent. We also document considerable heterogeneity in sponsorship across employer size, average annual salaries, and industry sector.
Using linked survey-administrative record data we find that among private sector wage and salary workers age 25-64, only 53 percent report that they work for an employer that offers any retirement plan (Defined Benefit or D.C.), while the administrative records indicate a D.C. offer rate of 75 percent. Among these workers, 44 percent report that they participated in a retirement plan, while the administrative records indicate a participation rate of 39 percent.
However, we emphasize that the currently available administrative records (1) cannot be usedto measure defined benefit participation, (2) can only be used to measure contributions totraditional (non-Roth) accounts and (3) do not measure D.C. plans where employers make contributions on behalf of employees even in the absence of an employee contribution. Thus, actual participation rates in any retirement plan are likely larger than we are able to observe at this time. Actual sponsorship rates may also be somewhat higher. In addition, some employeeswho work for an employer who offers retirement benefits may not themselves be eligible toparticipate in the plan. We also emphasize that more information about individual circumstances is required to draw firm conclusions about savings adequacy and retirement preparation. DRAFT
1
1. IntroductionAccuratemeasurementofemployer-providedpensioncoverageiscriticallyimportant
forunderstandinghowwellAmericansarepreparedforretirementandforconsidering
potentialpolicyreformstotheretirementsystem.YetresearchintheU.S.hasbeen
constrainedbydatalimitations.Inthispaper,wedevelopaninnovativeapproachtomeasuring
definedcontribution(DC)pensionplancoverageandparticipationusingtaxadministrative
records.Weexaminecharacteristicsofthismeasurebylinkingittoemployer-leveldatafrom
theCensusBureau’sBusinessRegister,plan-leveldatafromForm5500sfiledannuallywiththe
InternalRevenueService(IRS),andhousehold-leveldatafromtheCurrentPopulationSurvey
AnnualSocialandEconomicSupplement(CPSASEC)andtheAmericanCommunitySurvey
(ACS).Wepresentourmethodologyandresultsusingdatafrom2012.
UsingW-2taxrecordsfortheU.S.workforce,weidentifyfirmsthatsponsorDCplans
throughtheexistenceoftax-deferredretirementcontributionsbyanyemployeeofthefirmina
givenyear.Accordingtothismeasure,13.7percentofemployerssponsorplansand71.2
percentofjobsareatfirmssponsoringplans.Thesesponsorshipratesvarysubstantiallybyfirm
sizeandaverageemployeewages.Wealsolinkthenewemployer-levelmeasuretothe
BusinessRegister,allowingustoidentifyheterogeneityinsponsorshipandparticipationrates
andrelateittoarichsetoffirmcharacteristicsincludingindustryandlegalformof
organization.
Apotentiallimitationofthismethodologyisthatitdoesnotaddressdefinedbenefit
(DB)pensionsandprovideslittleinformationaboutcharacteristicsofplansidentifiedintheW-2
records.Toinvestigatethisissue,wefurtherlinktheW-2recordstotheForm5500dataand
estimatecoverageratesacrossbothDCandDBpensionplans,aswellasthepresenceofaDC
autoenrollmentplan.
BecausefewfirmsofferonlyDBplans,ourmeasureofDCcoverageusingW-2sbased
measureofDCsponsorshiproughlyapproximatesoverall(DBandDC)retirementplan
coverage.ThisanalysisalsoillustratestheimportanceofusingtheBusinessRegisterforfirms
thatusemultipleEINs.
DRAFT
2
Tounderscoretheimportanceofourmethodology,weusetheCPSASEClinkedtoW-2
recordstodocumentsubstantialdifferencesbetweentheW-2basedmeasureandsurveyself-
reportedpensionplancoverageandparticipation.Workersunderreportemployersponsorship
ofDCplansbyroughly20percent.Theseworkersareeitherunawarethattheiremployersoffer
pensionplans,ormisunderstandthesurveyquestion.This,inturn,leadstoestimatesoftake-
upratesthataresubstantiallylargerthanwhatwefindbasedonW-2coverageand
participation.
Wewerealsointerestedinexaminingcoverageratesacrossvariousdemographic
groupswithintheU.S.Todothis,weuseindividual-leveldatafromtheAmericanCommunity
Survey(ACS)linkedtotheW-2records,andweproducenationwideestimatesofretirement
plancoverageandparticipationratesforvariousdemographicsub-groups.Wefindlarge
differencesinparticipationandcontributionsacrosseducationandracegroups,evenafter
controllingforincome.
2. Background
2.1 PriorStudiesofPensionPlans
ThereisnowconvincingevidencethatDCplansponsorscanchangeparticipant
behaviorusingautomaticenrollment,defaultcontributionsandmatchrates(Papke,1995).
Chettyetal.(2014)arguethatroughly15percentofDanishworkersare“active”ratherthan
“passive”savers,meaningtheychangetheirDCcontributionswhenthereisadefault
contributionlevelthatdivergesfromtheiroptimallevel.
Inlightofthesefindings,policymakersareincreasinglyfocusedonincentivizingfirmsto
sponsorplansandprovidemoregenerousplancharacteristics.Forexample,theWhiteHouse
budgetfor2017includesprovisionsdesignedto“makeiteasierforemployerstocreatepooled
401(k)planstolowercostandburden.”Othercomponentsofthebudgetaimtoincrease
savingsratesatfirmsalreadyofferingDCplans,specificallyto“providetaxcutsforbusinesses
thatchoosetooffermoregenerousemployerplansorswitchtoautoenrollment”(TheWhite
House,2016).
DRAFT
3
Despitefirmsplayingacriticalroleinshapinghouseholdretirementsavings,andthe
increasingfocusinpolicydiscussions,littleisknownaboutfirm-leveldecisionsregarding
sponsorshipandplandesign.Researchershaveofferedexplanationsforwhyemployersoffer
pensionstotheirworkers,butthereisnoconsensusaboutwhichfactorsaremostimportant.
Previousliteraturefocusesonhowtheprimaryfunctionofpensionsistobondworkerstotheir
employers(Allen,Clark,andMcDermed1993;Ippolito1991).Someresearchfocusesonsupply-
sidefactorsinfluencingwhetherafirmoffersaplan,orwhethertoofferaDCordefined-
benefit(DB)plan(AaronsonandCoronado,2005;Dummann,2008),andmorerecentworkhas
investigatedwhetherDanishfirmstailorpensionplanstotheiremployees’savingpreferences
(Fadlonetal.(2016).
2.2 Dataondefinedcontributionsponsorshipandparticipation
IntheU.S.,themostcredibleresearchonworkerparticipationdecisionscomesfrom
responsestoexperimentalandquasi-experimentalvariationinplancharacteristics.These
studiesuseproprietaryplan-leveldatafromplansponsors(Nessmith,UtkusandYoung,2007;
VanDerhei,2010)anddatafromindividualfirms(Choietal.2004;MadrianandShea2001)to
exploitwithin-firmvariationovertime.Generalizingthesefindingstoaheterogeneous
populationisnotpossible.Thatwouldrequireeitherpopulation-leveldataorarepresentative
setoffirmsandworkers.
Plan-leveldatafromForm5500filingsprovidesinformationaboutspecificplans,buthas
severalshortcomingswhenstudyingfirmsponsorship,individualcoverageandindividual
participation.Form5500datadoesnothaveinformationaboutnon-sponsors.Second,some
sponsors—namely,publicsectorandsomesmallerSEPandSIMPLEplansponsors—arenot
requiredtofileForm5500.Anotherlimitationisthatpriorto2015,Form5500didnotcollect
informationonwhichemployersparticipatedinMultiemployerandMultipleEmployerplans.
TheNationalCompensationSurvey(NCS)fromtheBureauofLaborStatistics(BLS)
containsinformationaboutestablishmentsandtheirspecificplans,butlacksinformationabout
thecharacteristicsofworkerscoveredbytheseplans.Therelativelysmallsamplesizealso
limitsresearchers’abilitytostudysmallfirms,aswediscusslaterinthispaper.
DRAFT
4
HouseholdsurveyssuchastheCurrentPopulationSurvey(CPS),SurveyofIncomeand
ProgramParticipation(SIPP),andtheHealthandRetirementStudy(HRS)collectdatafrom
individualsincludingdemographiccharacteristicsandworkerself-reportsabouttheiremployer-
sponsoredretirementplans.However,workers’imperfectknowledgeabouttheirplans
(Mitchell,1988;Gustman,Steinmeier,andTabatabai,2009),highratesofsurveymisreporting
(DushiandIams,2010),andincreasingsurveynon-responserates(MeyerandSullivan,2015)
presentchallengeswhenanalyzingcoverageandtake-up.Inparticular,itseemsplausiblethat
workersunderreportcoverageinplansforwhichtheydonotparticipateandunobservable
characteristicsaffectingplanknowledgeaboutcoveragecouldbiasresultsconcerningtake-up.
WorkersmightnotknowaboutEmployeeRetirementIncomeSecurityAct(ERISA)rules
concerningnon-discrimination,and,therefore,donotparticipatebecausetheyassumethey
arenotcoveredundertheiremployer’splan.Ontheotherhand,workerswhodonotwantto
participateinapensionhavelittleincentivetoknowabouttheiremployer’splansponsorship.
Estimatesofnationwidepensioncoveragevarysubstantiallydependingonthesourceof
data,particularlywhethertheunitofanalysisistheworkerorthefirm.Additionally,high-
qualityworker-leveldataareneededtounderstandparticipation.Intherestofthispaper,we
showhowemployer-employeelinkeddataimproveourabilitytounderstandsponsorship,
participation,andthepotentialimpactofpoliciesincentivizingfirmsponsorship.
3. DataandmethodologyInordertounderstandthedeterminantsandconsequencesofemployer-sponsorship,
wedrawuponanarrayofadministrativerecordsanddevelopanewapproachtomeasuring
firmsponsorshipofDCplans.Inthissection,wedescribeindetailthedatausedforourstudy.
3.1 W-2Records
OurprimarydatasourceistheuniverseofW-2taxrecordsobtainedfromtheIRS.Each
year,employersissueaW-2toeachoftheiremployees.TheW-2scontaininformationon
annualearningsandemployeecontributionstoemployer-sponsoredtax-deferredDCplans
(e.g.,401(k)s,403(b)s.).ContributionsarebasedonamountsincludedinBox12oftheW-2.
DRAFT
5
WedefineannualearningsforeachjobasthesumofBox1(wages,salaries,bonuses,etc.)and
Box12amounts.
Wemeasureparticipation,sponsorshipandtake-upusingthisinformationfromtheW-
2s.Wedefineindividualparticipationashavingpositivetax-deferredDCcontributionsinthe
givenyear.Afirmsponsorsoroffersaplanifanyemployeeofthefirmparticipatesinagiven
year.Thetake-uprateistheshareofemployeesatthefirmwhoparticipate(amongfirmsthat
offeraplan).
Throughoutthispaper,thebasicunitofanalysisisthejob,whichisdefinedasan
employer-employeepair.Insomecases,weaggregateuptotheEmployerIdentification
Number(EIN)toproduceemployer-levelstatistics.Inothercases,weaggregateupfromthe
jobleveltotheworkerleveltomeasureretirementplanaccessratesofindividuals.Ineach,we
clearlyidentifytheunitofanalysisforourresults.WeusetheEINlistedontheW-2tolinkto
otherCensusemployerdatasources.EmployeesaretrackedbasedontheirProtected
IdentificationKey(PIK),whichisauniqueidentifierdevelopedbytheCensusBureauthat
facilitateslinkingindividualstootherindividual-leveldatasets.1
ThereareafewcaveatswithusingW-2recordstodefineDCsponsorshipand
participation.Whileitwouldbeworthwhiletoexaminedifferencesbetweenemployer-
sponsoredplansandothersponsors—suchasunions—wecannotdistinguishbetweenthetwo
inW-2records.Nevertheless,ourgoalistomeasureDCplancoverage,and,therefore,itdoes
notmatterwhethertheplanistechnicallysponsored(andmanaged)bytheemployerversus
otherarrangements.2AnotherchallengeisthatwemightnotobserveallDCplancontributions.
Itispossible(althoughuncommon)forafirmtoofferDCplanswithoutitshowinguponanyof
theiremployee’sW-2returns.Thiswouldhappenifemployerscontributetoplanswithoutany
employeecontributions.After-taxcontributionstoRoth-designatedaccountsarealsoabsent
fromourmeasureofcontributions.WhileparticipationinRoth-designatedemployer-sponsored
1SeeWagnerandLayne(2014)formoreinformationonthePIKassignmentprocess.2Thisdistinctionbetweenemployer-sponsoredandunion-sponsoredplansiscrucialwhenmergingW2recordstoinformationonplansinForm5500.WeaddresstheseissuesinthesectionabouttheForm5500data.
DRAFT
6
planshasgrownsincetheirintroductionin2006,firmsofferingtheRothoptiontypicallydoso
inadditionto(ratherthaninsteadof)tax-deferredaccounts.Infact,theemployercontribution
totheseaccountsmustbefrompre-taxearnings,requiringemployerandemployee
contributionstobemaintainedseparatelybyplansponsors.3IgnoringRothcontributionscould
leadtounderestimatingparticipationrates.Wedonotexpectthistoimpactoursponsorship
ratesforlargefirms,butcouldforsmallfirmsifallworkersmakeafter-taxratherthanbefore-
taxcontributions.
3.2 BusinessRegister
Weexploredifferencesinsponsorshipandparticipationratesacrossfirmcharacteristics
bylinkingfirmsintheW-2stotheU.S.CensusBureau’sBusinessRegister(BR)usingEINs.4This
databaseisdesignedtoincludeallactiveestablishmentsintheU.S.;itgeneratesthesampling
frameforestablishmentandfirm-levelsurveysfieldedbytheCensusBureau.WefocusonEINs
thatareintheBRrecordsin2012.EINsassociatedwithnon-employerfirmsaretypically
removedfromtheBRrecords.Single-unitfirmstypicallyhaveoneEIN,whereasmulti-unitfirms
mayhavemorethanoneEIN.TheBRhasinformationaboutwhichEINscorrespondtothesame
firm.ForsomeanalyseswecombineEINstoconstructafirmlevelmeasure.
3.3 Form5500
EmployersthatsponsoremployeebenefitsmustfileForm5500AnnualReturn/Report
ofEmployeeBenefitPlan(“Form5500”)formostbenefitplans.5Form5500helpscompanies
satisfyannualreportingrequirementsunderthe1974EmployeeRetirementIncomeSecurity
Act(ERISA)andundertheInternalRevenueCode.Becauseemployeesareentitledto
informationabouttheirpensionbenefits,thesefilingsaregenerallypublicrecords.The
3WearestillinvestigatingwhetherForm5500datahasinformationtoidentifyplanswithaRothoption.Sincetheemployerandemployeecontributionsmustbekeptseparate,it’spossibletheyshowupseparatelyinForm5500records.4TheBusinessRegister(BR)containsallEINswhoreportpayrollonIRSForm941or943inthecurrentyear.AlmostallEINsintheW-2recordsarealsointheBR.NewlyestablishedandsmallerfirmsarelesslikelytobeintheBR.SeeJarminandMiranda(2003)formoreinformationontheBR.5SmallerplansthatsatisfycertaincriteriamayinsteadfileaForm5500-SF,ShortFormAnnualReturn/ReportofEmployeeBenefitPlan(“Form5500-SF”).
DRAFT
7
return/reportmustbefiledwhetherornottheplanis“tax-qualified,”benefitsnolonger
accrue,contributionswerenotmadethisplanyear,orcontributionsarenolongermade.
SponsorsofbothDCandDBplansarerequiredtofile.Certainemployersandtypesofplansare
exemptfromForm5500filingrequirements.Thisincludesallgovernmentplans,andsomevery
smallprivatesectorplans,includingcertainSimplifiedEmployeePensions(SEPs),Savings
IncentiveMatchPlanforEmployeesofSmallEmployers(SIMPLEs),andone-participantplans
thatfile5500-EZ(whichisnottreatedaspublicinfo).
WeuseForm5500datafromthe2012ResearchFiletoenhanceourunderstandingof
estimatesofDCplansponsorshipusingW-2records.WhileW-2recordsmeasureDCcoverage,
wecannotmeasureoverallpensioncoveragewithoutinformationonDefinedBenefit(DB)
plans.TheForm5500dataalsoincludeplancharacteristicsthatareabsentintheW-2s.
4.ResultsTable1showsDCplancoverageandtake-upratesusingourEIN-levelmeasures
constructedfromtheuniverseofW-2recordsin2012.Overall,thereare6.2millionfirmsinthe
2012W-2records,andatleast13.7percentofthemsponsoredDCplans.Therewere218.9
millionuniquejobs,andatleast71.2percentwereatfirmssponsoringDCplans.Amongjobsat
thesefirms,33.6percentmadeDCcontributions.
[InsertTable1]
Sincepeoplecanhavemultiplejobs,wealsocalculateindividual-levelcoverageand
take-uprates.Amongthe154.8millionuniqueindividualswithaW-2,roughly78.5percent
haveoneormorejobswithDCcoverage.Thetake-uprateamongtheseworkersis41.4
percent.Restrictingtothe135.1millionworkerswithearningsequivalenttoparttime
employmentforhalftheyearatthefederalminimumwage($3,770totalannualearnings),
coverageis80.8percentandtake-upis45.7percent.6
6ThisrestrictionisthesameastheoneusedbySongetal.(2015).Itexcludesindividualswhodidnothaveareasonablystronglabormarketattachment.Weconsideranindividualtobefull-timeinagivenyearifsumming
DRAFT
8
3.3.1 DCsponsorshipandparticipation
Table2summarizessponsorshipratesandaverageparticipationratesacrossEINsize,
whereEINsizeisdefinedasthetotalnumberofjobsoccurringatsometimeduring2012.This
measureofEINsizeprovidesanupperboundonthenumberofemployeesattheEINata
particularpointintime.7PanelAreportsEIN-levelestimates,whereasPanelBreportsjob-level
estimates.InbothpanelsthefirstcolumnisthepercentageofEINs(jobs)ineachEINsize
category.Mostfirmsemployasmallnumberofemployees,yetmostemployeesworkatlarge
firms.In2012,over70percentoffirmshadfewerthan10employees,andover85percenthad
fewerthan25employees.Whileonlyone-thirdofonepercentoffirmshad1,000ormore
employees,theyaccountedforapproximately50percentofalljobs.
[InsertTable2]
Columnstwoandthreesummarizesponsorshipratesandaverageparticipationrates
acrossfirmsize.WefirstlookatthepercentageoffirmsofferingDCplans.Mostfirmsdonot
offerDCplans,butmostjobsareat(large)firmsthatofferplans.Thatis,only13.7percentof
firmsofferDCplans,buttheyaccountsfor71.2percentofthe218.9milliontotaljobs.While72
percentofEINswith100ormoreemployeesofferDCplans,only12percentofEINswith1-99
employeesofferthem.
Next,wecompareaveragetake-upratesamongemployeesatfirmsthatofferDCplans.
Theaverageparticipationrateacrossallfirmsis45percent.Take-upis100percentatsingle-
employeefirms,byconstruction.Take-upisover50percentforfirmswith2-9employees,and
41percentforfirmswith10-24employees.Theratethenplateausasfirmsizeincreases,
rangingfrom32and37percentforfirmswithatleast25employees.
Wealsogroupfirmsbynumberofemployees—1to9,10to99,and100ormore—and
computeanEIN-levelaverageofemployeeearnings.Firmsarethengroupedintoventilesby
acrossalljobs,heorsheearnsatleasttheequivalentof40hoursperweekfor13weeksatthefederalminimumwage($3,770in2012).7Thismeasureofsizeisdifferentfromthemeasureusedinmostbusinessstatistics,whichistypicallybasedonemployerreportsofnumberofemployeesatapointintime.
DRAFT
9
theaverageoftheiremployee’searnings.TheseventilesareconstructedwithinEINsize
categories.8
Figure1(A)plotsthemeanofferratesbyfirmsizecategoriesandacrossaverage
earningsventiles.Sponsorshipratesarehighestforlargefirmswithhighwages.The
sponsorshiprateisover98percentforthetopventileofemployeeearningsamonglargefirms.
SmallfirmsrarelyofferDCplans,andsponsorshipisconcentratedamongthoseinthetop
ventilesofaverageearnings.Thefigureshowsheterogeneityinsponsorshipwithinfirmsize
categories.Therearetwointerestingpatterns.Sponsorshipratesarehigherforsmallfirmsin
thetopventileofearnings(27percent)thanforlargefirmsinthelowestventileofearnings(14
percent).Withineachcategoryoffirmsize,sponsorshipratesincreasewithmeanemployee
earnings.Thesechangesinsponsorshipratesareatlowermeanearningsamonglargefirms
thanamongsmallfirms.
[InsertFigure1]
Afirm’slikelihoodofofferingDCplansdependsonthedemandfortheseretirement
benefits.Itisonlyworthwhileforanemployertosponsorplansiftheyexpectadequate
participationamongcoveredemployees.Thesepatternsofsponsorshipacrossfirmsizeand
meanemployeeearningscouldbeexplainedbysponsorshipdecisionsdrivenbyexpected
numberofparticipants.Individualparticipationratesincreasewithearnings—asweexamine
laterinthepaper—andthereforeexpectedparticipationincreasesinthenumberofworkers
withhighearnings.
Figure1(B)plotsaverageparticipationratesforfirmsgroupedbysizeandaverage
employeeearnings.AsevidentinTable2,smallerfirmshavehigherparticipationrates.ForEINs
with100ormoreemployees,averageparticipationratesaremonotonicallyincreasingin
averagetotalearnings.Averageparticipationratesincreasemonotonicallyamonglargefirms.
8ItwouldbeclearertousethesameearningsthresholdstogroupEINs,regardlessofEINsize.Resultsinvolvingothercut-offsandusingmediansinsteadofmeansaresubjecttodisclosurereview.
DRAFT
10
Smallandmediumsizedfirmsillustrateadifferentpattern.Participationratesarehighestat
thebottomandtopendofthemeanearningsdistribution.9
3.3.2 Individualparticipationacrossincome
Theseresultssuggestthatsponsorshipandparticipationratesvaryacrossaveragefirm
earnings,particularlyatsmallandmedium-sizedfirms.Tobetterunderstandtheseresults,
Table3showshowcoverageandparticipationratesvaryacrossworkerearnings.Table3shows
jobswithlowerearningsarelesslikelytobecoveredbyDCplans,atfirmssponsoringDCplans.
Thereareevenlargerdifferencesinparticipationrates,andtake-upratesvarysubstantially
acrossincome.Amongworkersearningbetween$100,000and$150,000in2012,roughly93
percentwereatfirmssponsoringDCplansand76percentmadetax-deferredcontributions,for
atake-uprateof82percent.Incontrast,amongworkersearning$20,000to$30,000in2012,
roughly75percentwerecoveredbyDCplansyetonly31percentparticipated,foratake-up
rateof41percent.Take-uprateswereevenloweratjobsearninglessthan$20,000.
[InsertTable3here]
Therelationshipbetweentake-upandincomeisconsistentwithsponsorshiprates
acrossfirmsizeandaveragefirmearnings.Firmsneedsufficientworkerparticipationtooffset
costsassociatedwithsponsoringDCplans,andsufficientworkerparticipationonlyhappens
amonghighlypaidemployees.Inlightofthetake-upratesacrossearnings,wewouldexpect
lowsponsorshipratesatsmallandmediumfirmsthatpaylowwages,asillustratedinFigure1.
3.3.3 TenurerequirementsforDCcoverage
WehavethusfarimplicitlyassumedallworkersatfirmssponsoringDCplansareeligible
toparticipate.WhilenondiscriminationrulesunderERISAconstrainfactorsthatlimitemployee
eligibility,somerequirementsstillexist.Ourjob-levelestimatesignoretheseeligibility
requirements.Asaresult,ourjob-levelestimatesarelikelyanupper-boundoncoveragerates
andalowerboundonparticipationrates(take-up).
9Theremightbeamechanicalrelationshipamongsmallfirmswithlowermeanwages.Iffirmswithoneemployeeareinthelowestventilesofearnings,thentheparticipationrateismechanicallyhigh.Forthesefirmstheaverageparticipationrateisone,sincethisfocusesonfirmsthatofferDCplans.
DRAFT
11
Employeesgenerallymustbeallowedtoparticipateinaqualifiedretirementplanifthey
areatleast21yearsoldandhave1yearofservice.Therearesomesituationsinwhichworkers
mightnotbeeligibleforemployercontributionsuntil2yearsofservice.However,employees
mustbeallowedtomakeelectivedeferralcontributionsafter1yearorless.10
Toinvestigatethepotentialroleoftenurerequirements,wetakethesubsetoffirmsin
2012thathave1styearemployeesand“tenured”employees.Inparticular,weanalyzeEINsthat
existfor2010,2011and2012,withatleastonefirst-yearemployee(in2012)andatleastone
“tenured”employee,whichwedefineasbeingintheW-2recordsin2010or2011inaddition
to2012.
Table4presentsDCplancoverageratesforthissubsetofEINs.While27.8percentof
thesefirmsofferDCplansin2012,only9.3percentofthemhaveplanswithfirst-year
employees.Thissuggeststhatupto18.5percentoffirmshavetenurerequirements,whichis
aroundtwo-thirdsoffirmsofferingDCplans.Forfirmswithfewerthan10employees,8.2
percentofferDCplans,andonly1.1havefirstyearemployeesmakingcontributions.In
contrast,amongfirmswith100ormoreemployees,74.2percentofferplans,with45.2offering
planstofirst-yearemployees.Thedifference39.9percentisanestimateoftheupperboundon
theshareoffirmsthathavetenurerequirements.Theseresultsareconsistentwithsmaller
firmsbeingmorelikelytohavetenurerequirements.Thiscouldbeexplainedbysmallerfirms
havinggreaterincentivetoretainemployeesduetohigherhiringandtrainingcostsperworker.
However,thereisalsoamechanicalrelationshipwithlargerfirmsbeingmorelikelytohave
bothtenuredandnewemployees,andalsotohavepeoplecontributingatall.
[InsertTable4]
3.4 HeterogeneityinsponsorshipbyBusinessRegisterfirmcharacteristics
Table5presentssponsorshipratesandaverageparticipationacrossfirmtypeand
primaryindustry,highlightingsubstantialheterogeneityacrossfirms.Employer-levelmeasures
10Thisrequirementdoesnotapplytocertain“leased”employees.
DRAFT
12
ofsponsorshipandwithin-firmtake-upratesarethenlinkedtofirmcharacteristicsfromthe
BusinessRegister(BR).Categoriesoforganizationincludenon-profit,federalgovernment,
state/localgovernment,corporation,soleproprietorship,andpartnership.Wegroupfirmsby
industryusingtheNorthAmericanIndustryClassificationSystem(NAICS)Sectorcodes.
[InsertTable5]
PanelApresentsofferratesandaverageparticipationratesacrossfirmtype.Not
surprisingly,only3.3percentofsoleproprietorshaveDCplans.11
Thisnumberislargerthanthenumberofone-employeefirmswithDCplansforacouple
reasons.Peoplewhoarehiredassomeone’sloneemployeearelikelydifferentthansomeone
whoisself-employed(soleproprietor).
ThehighestratesofsponsorshipareamongState/localgovernments,withoverhalfofferingDC
plans,followedbynon-profitorganizationsat28percent.Amongfor-profitbusinesses,12.7
percentofS-corporations(S-corps)and14.2percentofPartnershipsofferDCplans,compared
to19percentofC-corporations(C-corps).12ComparedtoS-corps,C-corpstendtobeolderfirms
employingoldertenuredworkers.S-corpsandpartnershipsarebothpassthroughentities—the
company’sprofitsshowupasincomeontheindividualreturns.
Thesecondcolumnshowstheaverageparticipationratesacrossfirms.Theaverageis
overthepercentparticipatingateachfirm.Thelowestaverageparticipationrateisamongnon-
profitsandstate/localgovernments,around30percent.State/localgovernmentworkersare
morelikelytohaveaccesstodefinedbenefit(DB)plans,aswell,whichcouldexplainthelower
averageDCparticipationrates.ThehighestaverageparticipationratesareforSandC-corps,
botharound49percent.Theaverageparticipationrateforsoleproprietorsislessthan100
11SoleproprieterreferstoabusinessthatfiledScheduleC,soitisnotanincorporatedbusiness.Whilemostdonothaveemployees,theycan.Businesseswithemployeeshavealotofincentivetoincorporate,becauseotherwisethefaceliabilityissues.Iftheyincorporate,theymaytreatthemselvesasanemployee.12AnS-corporationisaspecialtypeofcorporationwiththeSubchapterSdesignationfromtheIRS.AnimportantdifferencebetweenC-corpsandS-corpsisthatprofitsandlossesfromS-corpsgetpassedontopersonaltaxreturnsandthebusinessesthemselvesarenottaxed.ThereisalsomorelimitedliabilityprotectionforS-corps,andshareholdersarenotnecessaryshieldedfromlitigationfromemployees.
DRAFT
13
percentbecausesoleproprietorshipisabouttheownershipstructureratherthanthenumber
ofemployees.
PanelBpresentsthesamestatisticsbutgroupingfirmsbymajorindustrysector.
Utilities,managementandpublicadministrationhadthehighestratesofDCplansponsorship,
whereasthelowestrateswereinaccommodationandfood,andagriculture,forestryand
fishing.Construction(9.5percent)andtransportation(10.7percent)arealsoindustrieswith
manyfirmsandofferratesthatarebelowtheoverallmeanofaround14percent.Overall,
industrieswithsmallerfirmsandhigherturnoverhavehigherratesofDCsponsorship.Average
participationratesarehighestamongutilities,management,andthefinanceandinsurance
sectors,allover60percent.Participationratesarealsohighforfirmsintheprofessional
services(59percent)andmining(56percent)sectors.Theyarelowestintheaccommodation
andfoodsector(13percent),whichoftenemploysmanypart-timeandshort-termworkers.13
3.5 CoverageinpensionplansusingForm5500
Next,wecompareourestimatesofDCcoverageusingW-2swithbothDCandDB
sponsorshipinForm5500data.WefocusonprivatesectorEINsthatareintheW-2sandthe
BusinessRegister.14Publicsectorjobs(federal,stateandlocal,orIndiantribalgovernments)
areexcludedfromthesamplebecausetheyareexemptfromForm5500filingrequirements.
InTable6,eachEINiscategorizedbywhetheritofferedDCplansinW-2s(in2012),and
bywhetheritofferedDCorDBplansaccordingtoForm5500(planyearendingdateincalendar
year2012).AnEINofferedaDCplanaccordingtoF5500iftherewasoneormoreDCplans
associatedwiththeEINwithactiveparticipants.AnEINofferedaDBplaniftherewasatleast
oneDBplanwithactiveparticipants.13Thesenumbersarepurelydescriptive.Thereareknowndifferencesinfirmsizeacrossindustry,andwehaveshownsponsorshiptoincreaseinfirmsize.Therearealsodifferencesinemployeeturnoveracrossindustry,whichcouldimpactparticipationrates,particularlyatfirmswithtenurerequirementsforDCplanparticipationandvestingofemployercontributions.Certain“leased”employeesarealsoexemptfromnondiscriminationrulesunderERISA.14MostEINsintheF5500dataareassociatedwithfirmsintheBR.WealsocheckedwhetherEINsareassociatedwithnon-employerfirms.AmajorityofEINsthatareinForm5500butnotintheBRareindeednon-employerfirms.Infutureworkwemayincorporatetheseintotheanalysis,butfornowweareonlylookingatEINsthatlinktotheBR.
DRAFT
14
ThefirstpanelofTable6—“UsingEIN-levelmeasures"—focusesonmeasuresof
sponsorshipinW-2recordsandForm5500attheEIN-level.ThecolumnforEINsgivesthe
distribution(inpercent)ofEINsthatdonotofferDCorDBplansinW-2orForm5500(84.4
percent),offerDCinW-2swithnocorrespondingF5500plans(6.5percent),andsoforth.15The
columnfor"Jobs"isthedistributionofjobs(inpercent)atEINsthatofferplans:e.g.,31.6
percentofjobsareatEINsthatdonotofferplans.AlittleoverhalfofthejobsatEINsthatoffer
DCintheW-2s(66.8percentofjobs=30.2+28.5+0.7+7.4)haveF5500recordsforDCplans
(35.9=28.5+7.4).ThepercentagesofjobsarecomputedbyweightingeachEINbythenumber
ofdistinctworkerswithW-2recordsin2012.
[InsertTable6]
ThereareafewreasonswhymanyEINsintheW-2sdonothaverecordsinF5500.Some
verysmallplans(SEPs,SIMPLEs,andsingle-employeeplansthatfile5500-EZ)areexemptfrom
F5500reportingrequirements.Informationonsuchsmallplansisnottreatedaspublic
informationandthereforeunavailableintheF5500researchfiles.TherecouldbemultipleEINs
thatareownedbythesamebusinessentity.Thatis,therecouldbetwoEINsusedforpayroll,
buttheF5500filingisonlyreportedusingoneofthem.Alternatively,therecouldbeaseparate
EINforpayrollandforF5500filings.Finally,multiemployerandmultipleemployerplanswere
notrequiredtoreportallEINsassociatedwiththeseplansuntil2015.
TheissueofmultipleEINscanbedealtwithbyleveragingtheCensusBureau’sBusiness
RegistertolinkEINsthatareassociatedwiththesamefirm.Wepresentnumbersbasedonthis
linkageinPanelB.—“Usingfirm-levelmeasures."Morespecifically,W-2andF5500information
collapsedtothefirmlevelusingthefirm-levelidentificationvariablestogroupEINsformulti-
unitfirms.Here,afirmoffersaDCplaninW-2sifanyEINassociatedwiththefirmoffersaDC
plan.Similarly,afirmoffersaDC(orDB)planinForm5500ifanyEINassociatedwiththefirm
sponsorsatleastoneDC(orDB)planintheForm5500.Thedistributionoffirmsacross
categoriesofDCandDBplancoverageissimilartothedistributionofEINsacrossthese
15Forexample,therow“DC(W2,5500);DB(5500)”meansanEINofferedaDCplaninbothW-2andForm5500,andofferedaDBplaninForm5500.
DRAFT
15
categories.Thedistributionofjobsacrosscategorieschangesalot.Thepercentageof“none”
decreasedslightlybecausetherewereEINsassociatedwiththesamefirmthathadDC
contributionsassociatedwithoneEINandnoDCcontributionsassociatedwithanotherEIN,
andyetneitherEINshowedupintheForm5500.ThenumberofjobsatfirmswithDCplans
accordingtobothW-2sandF5500increasedfrom28.5percentto38.1percent,andtheshare
withDCintheW-2onlydecreasedfrom30.2percentto11.4percent.Thisfindingisitself
important:mostjobswithDCplansintheW-2appearinForm5500records,aftergrouping
EINstogetherusingtheBRinfrastructure.Becauseofthis,wecanuse5500datatoassignplan
characteristicstoW-2records,aswedescribenextinthecontextofautoenrollment.
Nevertheless,thenumbersthatarefromcollapsingbytheBRidentificationvariablesshouldbe
thoughtofasanupperboundonthejobs-levelcoveragerates.
Finally,wenotethatonly0.1percentofEINsofferonlyDBplans,withroughly0.1
percentofjobsatsuchfirms.IntermsofidentifyingwhetherornotanEINofferspensionplans
(DBorDC),ourW-2basedmeasureofDCsponsorshipcloselyapproximatesoverallpension
coverage.
3.6 Autoenrollment
AnothershortcomingofmeasuringDCsponsorshipusingW-2recordsaloneisthat
importantplancharacteristicsarelacking.TheForm5500datahelpaddressthisissue,
particularlyinmeasuringemployeradoptionofautoenrollment.
WeassignplancharacteristicstofirmsusingtheForm5500andtheBusinessRegister
infrastructure.Table7presentsestimatesofDCplanautoenrollmentcoverage.EINsare
categorizedintothefollowingrows:(1)DCcoverageinW-2s,autoenrollmentaccordingto
F5500.(2)DCcoverageinW-2s,noautoenrollmentaccordingtoF5500.(3)DCcoverageinW-
2s,unknownautoenrollmentbecausenoactiveDCplansassocatedwiththisEINinF5500
records.(4)NoDCcoverageinW-2s.
[InsertTable7]
DRAFT
16
ThecolumnsareconstructedsimilarlytoTable7.Thefirstcolumnisthedistributionof
EINsbywhatweknowaboutDCplanautoenrollment.Only0.4percentofEINsofferDC(inW-
2s)anduseautoenrollment(accordingtoForm5500filings),6.9percentofEINsofferDCinW-
2s(andfileForm5500)butwithoutautoenrollment,and6.5percentofEINsofferDC(inW-2s)
butwithunknownautoenrollmentstatusbecausethereisnocorrespondingForm5500record.
AsinTable1,86.2percentofEINsdoNOTofferDCplansinW-2s.(NOTE:tocomparethiswith
thepreceedingtable,thisincludes84.4+1.6+0.1+0.1=86.2).Thecolumnfor"Jobs"isthe
distributionofjobs(inpercent)atEINsthathaveautoenrollment(6.3percent),donothave
autoenrollment(29.5percent),haveunknownautoenrollment(30.9percent)orhavenoDC
coverage(33.2percent).
Thesecondpanel—“Usingfirm-levelmeasures”—usesallW-2informationfromthe
EINsandForm5500informationassociatedwiththefirm.IftherearenoForm5500records
associatedwiththeEIN,weappendinformationontheactiveDCplanssponsoredbyotherEINs
associatedwiththesamefirm.Here,anEINhasDCautoenrollmentif(a)aForm5500record
existsfortheEIN,withautoenrollment,(b)noForm5500recordexistsfortheEIN,butthereis
leastoneactiveDCplanwithautoenrollmentatotherEINsassociatedwiththesamefirm.
ThereisDCcoveragewithunknownautoenrollmentiftheEINoffersDCplaninW-2,butthere
isnoactiveDCplanassociatedwiththefirminForm5500records.Therearestill86.2percent
ofEINswithnoDCcoveragebecausethisfocusesonEINsofferingDCinW-2s,ratherthan
changingtofirmsofferingDCinW-2s.Thejobs-leveldistributionchangesmorethantheEIN-
leveldistribution.13.4percentofjobsareatfirmswithautoenrollment,and41.8percentareat
firmsthathavehavenoplanswithautoenrollment.
4. Comparingnewmeasurestoself-reportedratesfromCPSTobetterunderstandthestrengthsandweaknessesofmeasuringDCplansponsorship
andindividualparticipationusingW-2records,wedirectlycomparethesenewmeasuresto
self-reportedinformationfromtheCurrentPopulationSurveyAnnualSocialandEconomic
DRAFT
17
Supplement(CPSASEC).16Eachyear,betweenFebruaryandApril,theCPSASECsurveys
approximately75,000householdsabouttheirincomesandcharacteristicsoftheirjobsheldin
previouscalendaryear,includingtwoquestionsaboutemployment-basedretirementplans.
First,thosewhoworkedinthepreviouscalendaryearareasked:
OtherthanSocialSecuritydid(ANY)employerorunionthat(name/you)workedforin(thepreviouscalendaryear)haveapensionorothertypeofretirementplanforanyofitsemployees?
Iftheyanswer“yes,”theyarethenasked:
(Were/Was)(name/you)includedinthatplan?
Anaffirmativeresponsetothefirstquestionindicatesworkingatanemployerthat
sponsoredaretirementplanandanaffirmativeresponsetoboththefirstandsecondquestions
indicatesparticipatingintheplan.Whilethesemeasuresareregularlyusedtomonitor
retirementplancoverageandparticipation,Copeland(2015)showsthatCPSASECcoverage
andparticipationratesdivergefrompatternsintheNationalCompensationSurvey(NCS)and
elsewhere,raisingconcernsabouttheefficacyofusingCPStostudyretirementplancoverage.17
ThefirstCPSquestionisdesignedtoidentifycoveragebyDCorDBplansatanyjob
duringtheyear.18IntheW-2records,theequivalentforDCcoveragemeanshavinganyjobin
theyearatafirmthatoffersaplan.Similarly,intheW-2recordsaworkerparticipatesifshe
makesDCcontributionsatanyjobinagivencalendaryear.Conceptually,thiscorrespondswell
withaffirmativeresponsestobothCPSquestions,exceptDBisabsentintheW-2s.Throughout,
theCPSASECsampleisrestrictedtoindividualsage25to64whoreportworkingatawageand
16InthisversionweareonlycomparingCPSresponsestomeasuresofDCcoveragefromtheW-2records.WeplantobringinForm5500informationonDBcoverage,butourabilitytodosoislimitedbyexistingdatauseagreements.17WhileCopelandfocusesmainlyonchangesthatoccurfollowingtheCPSASECredesignin2014(about2013,heraiseesconcernsabouttheefficacyofthequestionsoverall.18ThereareconsiderableadvantagestotheCPSASECapproachtomeasuringretirementplancoverage.Evenifrespondentsdonotknowthespecifictypeofplansoffered,theycanstillanswerthatthereexistssometypeofretirementplanconnectedtoatleastoneoftheirjobs.
DRAFT
18
salaryjobinthepreviouscalendaryear(2012).Werestrictouranalysestopersonswhoweare
abletolinktoadministrativerecordsviathePIK.19
Table8summarizestheagreementof2013CPS-ASECand2012W-2retirementplan
information,forpensioncoverageandparticipation.Lookingfirstatpensioncoverage,roughly
47percentofrespondentssaidtheywerecoveredbyaplanandworkedatfirmsofferingplans
accordingtotheW-2s.Another27.7percentreportthattheiremployerdoesnotofferany
pensionplanseventhoughtheW-2recordsindicateotherwise.AllCPSrespondentsatfirms
thatofferDCplansintheW-2recordsshouldhaveanswered“yes”tothefirstquestion.Only
5.7percentofrespondentshaveaCPSmeasureindicatingDCsponsorship,butwithout
evidenceofDCsponsorshipaccordingtotheW-2s.Errorratesareslightlylargerforcaseswith
imputedresponses,althoughtheyaregenerallysimilartothenon-imputedcases.
[InsertTable8]
ComparingparticipationratesinCPSandW-2sprovidesevidenceoferrorsinCPSaswell
aslimitationsofusingW-2records.Overall,72.6percentofworkersareconsistentacrossthe
twomeasures—27.7percentparticipateinbothCPSandW-2,and44.9percentofworkers
participateinneither.Thereisalsoevidenceofdisagreementbetweenparticipationreportedin
CPSandwhatisobservedinW-2records.Thisincludes11.6percentofworkerswhohaveDC
contributionsintheirW-2sbutreportinCPSthattheydonotparticipate.
Therearealso15.8percentofworkerswhoreportparticipatinginCPSanddonothave
DCcontributionsintheirW-2s.Inthesecases,wecannotconcludethatCPSrespondents
misreportpensionparticipation.First,usingCPSwecannotdistinguishbetweenDBandDC
sponsorshipandparticipation.SinceonlyemployeeDCcontributionsarerecordedintheW-2
records,ouradministrativerecordmeasuremayfalselylabelworkersasnotparticipatingornot
beingofferedaplaneventhoughtheyhaveaDBplan,DCplanwithonlyemployer
contributions,and/oraplanrelatedtoaunionthatisnotrecordedontheW-2.Thereisalso
19Throughout,weobserveaclosecorrespondencebetweenthefullCPSASECandthePIKSamplesuggestingnotmuchislostbyexaminingonlyresponsesforthosethatwereassignedaPIK.OurdiscussionoftheCPSASECwillthereforefocusonthePIKSample.
DRAFT
19
ambiguityintheCPSwording,asitisunclearwhatitmeanstobe“included”inaretirement
plan.Aworkercouldhavea401(k),butmakenocontributiontoitinaparticularyear.Some
CPSrespondentsmaystillconsiderthemselvesincludedinthat401(k)planeveniftheyarenot
currentlycontributing.Again,thisshouldmaketheCPSmeasurelargerthanourW-2measure.
ThesediscrepanciesinthedatacouldbeexplainedbydifferencesinwhatwemeasureinCPS
andW-2s,andillustratesomelimitationsofusingW-2stomeasureparticipation.
Despitetheseconceptualdifferencesthatshouldleadtohigherreportedsponsorship
andparticipationinCPS,ourW-2measureofsponsorshipisconsiderablyhigherthantheCPS
measure,andourparticipationmeasureisonlyslightlylowerthantheCPSmeasureforprivate-
sectorworkers.Thissuggeststhat“true”coveragemightbeevenhigherthanourW-2measure
indicates.Ifwewereabletoincluderetirementplaninformationthatwecannotobserveinthe
W-2s,trueparticipationratesmightalsobehigherthantheCPSsuggests.
Table9comparesCPSresponsesandW-2measuresofretirementplancoverage,
participation,andtake-upratesfor2012.Weseeimportantdifferencesacrossdemographic
sub-groups,whichareattributablesolelytodifferencesbetweenthesurveyandadministrative
retirementplanmeasures.
Forworkerswhoreportedthattheirlongestjoblastyearwasaprivatesector,wageand
salaryposition,theCPSsponsorshipratewas53percent,whileitwas75percentintheW-2s.
Therewasalsoadifferenceinofferratesforpublicsectorworkers--81percentintheCPSASEC
versus94percentintheW-2s.Theseresultsareconsistentwiththeideathatsurvey
respondentsarefrequentlyunawareofwhattheiremployeroffersparticularlyiftheyarenot
themselvesparticipatingintheplan.
Turningtoparticipationrates,weobserveadifferentpattern.Forprivate-sector
workers,theW-2participationrateisabitlowerat39percentthantheCPSrateof44percent.
BecauseourW-2offerratesaremuchhigherandtheparticipationrateissomewhatlowerfor
theprivatesector,take-upratesasmeasuredbytheW-2sarealsomuchlower—53percent
versus83percent.
DRAFT
20
Forpublic-sectorworkers,theparticipationdifferencesaremorestarkat43percentin
theW-2sversus74percentintheCPS.Thiscouldbeattributabletothelackofdefinedbenefit
informationavailableintheW-2records.TheNCSdemonstratesthatalmostallstateandlocal
workershavearetirementplanandthattheCPSunderstatesbothofferandparticipationrates
inthepublicsector.Therefore,wefocusonlyontheprivatesectorworkersfortherestofthe
results.
TheCPSsuggestsprivatesectorsponsorshipratesrisewithage,from47percent,for
thoseage25to34,to56percentforthoseage55to64.Incontrast,theW-2sshowthat
sponsorshipratesaremuchhigherandflatteracrossagegroups,withaconsistentrateof
approximately75percent.Thereismoreagreementbetweenthetwosourceswhenitcomes
toparticipationrates.Theyrisefrom35percentto48percentintheCPSASECand32percent
to44percentintheW-2s.
Sponsorshipratesdovaryinbothdatasourcesbyeducationalattainment.Theyrange
from26percentforthosewithlessthanahighschooldegreeto64percentforthosewitha
collegedegreeormoreintheCPS.IntheW-2sthecorrespondingrangeisfrom55percentto
82percent.Onceagain,theparticipationratesmorecloselycorrespondbetweenthetwodata
sourcesrangingfrom17percentto56percentintheCPSandfrom16percentto53percentin
theW-2s.
Retirementplanpatternsbyracearealsonoteworthy.Inparticular,theCPSsuggests
sponsorshipratesarethelowestforblacksat48percentwhiletheW-2ssuggesttheyare
actuallythehighestat79percent.Intermsofparticipationrates,blackparticipationislower
thannon-HispanicwhiteparticipationinboththeCPSandtheW-2s.Hispanicsofanyracehave
thelowestparticipationrateat25percentandalsosomewhatlowersponsorshipratesat64
percent.
DRAFT
21
Resultsbyfirmsizeshowthatsurveyrespondentsatboththesmallestfirmsandlargest
firmsunderestimatefirmsponsorshipbyroughly20percent.20Becausesmallerfirmsareless
likelytoofferretirementplans,sponsorshipunderreportingislargerinproportionaltermsfor
thesmallestfirms.Participationratesareapproximatelythreetofivepercentagepointslower
intheW-2sthanintheCPSregardlessoffirmsize.
ForapproximatelyonequarteroftheCPSsample,unioncoveragestatuscanbe
determined.21Asisthecaseforthefullsampleofprivatesectorwageandsalaryworkers,
sponsorshipratesareunderestimatedregardlessofunionstatus.However,discrepanciesin
participationratessuggesttheCPSdoesabetterjobinmeasuringretirementplanparticipation
forunionworkersthantheW-2.Specifically,forunionworkers,theparticipationrateis66
percentusingtheCPSbutonly43percentintheW-2;theanalogousnumbersfornon-union
workersare43percentand42percent,respectivelyTheremainingprivatesectorprevalence
ofDBplansforunionworkersexplainswhywecannotreliablymeasuretheirretirementplan
participationintheW-2salone,giventhatonlyDCcontributionsarerecordedinthoserecords.
Giventhatprivatesectorunionizationratesarenowbelowsevenpercent,theunderstatement
ofoverallparticipationratesduetoDBpensionsislikelylimited.
Wehavesofarcomparedsurveyandadministrativerecordestimatesforasingle
calendaryear.BylinkingseveralyearsofCPSandW-2srecords,wecanalsoexaminehowthe
twodatasourcescompareovertime.Figure2(A)showsthatbetween2005and2012,
sponsorshipratesremainroughlysteadyintheW-2swhiletheyactuallyshowadeclineof
nearly5percentagepointsintheCPS.Intermsofparticipation,Figure2(B)showsthat
participationrateshavetrendedupwardintheW-2sfrom36.9percentto39.3percentwhile
theyhavedeclinedfrom46.3percentto43.5percentintheCPS.Figure2(C)showssimilar
patternsfortake-uprates,althoughthesechangesaresmallerthanforcoverageand
participationrates.BecauseparticipationratesforCPSincludebotherDBandDCplans,andfor
20TheCPSASECasksindividualstoreportthetotalsizeoftheiremployeracrossalllocationsfortheirlongestjobheldinthepreviouscalendaryear.21Thosehouseholdswhicharepartofthe“outgoingrotations”areaskedadditionalquestionsafterthebasicCPSinterviewincludingwhetherornotaworkerisaunionmemberorcoveredbyaunioncontract.
DRAFT
22
W-2sitisonlyDCplans,itcouldbethatDBparticipation(andcoverage)isdecreasingeven
morethanDCisincreasing.Wethereforecannotconcludethaterrorsareincreasingovertime.
Inshort,theaboveanalysishasshownthattheW-2shavemanyadvantagesin
measuringretirementplansponsorshipratesoversurveys.Whenitcomestoparticipation,
however,theresultsaremoremixed.Thepoint-in-timeestimatesareslightlylowerintheW-
2sthantheCPSASECin2012(withtheexceptionofunionandpublicsectorworkers)butthe
trendsaremovinginoppositedirections.
5. IndividualcoverageandparticipationWecanalsoexplorecoverageandtake-upacrossdemographicgroupsusingAmerican
CommunitySurvey(ACS)datalinkedtoW-2records.22Weusetheemployersponsorshipand
individualparticipationmeasuresderivedfromW-2records,andinformationoneducational
attainmentandraceandethnicityfromtheACS.ACSdataarematchedtoW-2recordsatthe
person-levelusingPIKs.
DatafromACSalsoallowsustobettermeasurecoverageamongthepopulationoffull-
timeemployeeswhoshouldbecoveredbyERISAnon-discriminationrules.Theanalysissample
includes2012ACSrespondentswithvalidPIKswhosatisfythefollowingsamplerestrictions:(i)
haveatleast1W-2recordwithstrictlypositiveearningsin2012,andhaveaggregateearnings
under250,00023;(ii)ages25through64;(iii)reportworking27ormoreweekstheprevious
year,andworkedanaverageof20ormorehoursperweek;and,(iv)haveamainjobinthe
privatesector.24
22FormoreinformationabouttheACSsampledesignandothertopics,visit<www.census.gov/acs>.Formoreinformationondataaccuracyandsamplingerror,pleasesee<https://www.census.gov/programs-surveys/acs/technical-documentation/code-lists.html>.23Wealsoremovedcaseswithover5W-2recordsfrom2012becauseitisdifficulttoidentifyamainemployer;thesecouldbecontingent/leasedemployeeswhoarenotprotectedunderERISAnon-discriminationrules.24Weidentifya“mainjob”usingtheW-2recordsonly,anddeterminewhetheritisprivateorpublic-sectorusinginformationfromtheBR.WerestrictfirmsusingthesamecriteriaweusedtoanalyzeW-2andForm5500coveragerates.
DRAFT
23
Table10presentscoverage,participation,take-upandcontributionratesforourACS
sampleofover1.3milliontenuredfull-timeprivate-sectorworkerswhoareeligibleto
participateinDCplansofferedbytheirmainemployer.AmongthissampleofACSrespondents,
75.4percentwerecoveredbyaDCplanattheirmainemployerand43.4percentparticipated
intheplan,foratake-uprateof57.7percent(43.4/75.4).Amongthosewhocontributedto
theseplans,employeecontributionswereanaverageof6.9percentofearnings(atthemain
job).WithinoursampleofACSrespondents,coverageandparticipationratesdifferacross
educationalattainmentandrace/ethnicity.
5.1 Heterogeneityacrossworkercharacteristics
Differencesbyraceandeducationalattainmentcouldstemfromdifferencesinthe
earningsdistributionacrossgroups,sincecoverageandparticipationvarybyincome(Table3).
Wenowexploredifferencesacrosseducationandraceafterholdingincomeconstant.We
measureraceusingfivecategories:(i)White,non-Hispanic.(ii)Black(Hispanicandnon-
Hispanic);(iii)Asian(HispanicandNon-Hispanic);(iv)Hispanic(notBlackorAsian);(v)Other
(whichincludesACScategoriesforAlaskanandAmericanNative,NH/OPI,andOther).
Educationalattainmentreferstoself-reportedhighestdegreeattainedatthetimeofthe
survey:(i)lessthanahighschooldiploma;(ii)highschooldiplomaorGED;(iii)somecollege;(iv)
four-yearcollegegraduate;(v)graduatedegree.
Forillustrativepurposes,wepresentgraphicalanalyses.ACSrespondentsarefirst
placedintodecilesofincome,usingthesumoftotalearningsacrossalljobsintheW-2
records.25Weplotestimatesofcoverage,participation,take-upandcontributionratesacross
incomedecilesseparatelybyraceandeducation.26Thefiguresillustratewithinandacross-
incomedifferencesbyraceandeducation.Whilethesamplesizesarequitelarge(hundredsto
thousandsforeverysub-sample),thedifferencesacrossdemographicgroupsareforillustrative
25Themeanincomeforthetwogroupsarenotnecessarilyequal,asisapparentforthetopbininsomeofthefigures.26Wefocusonraceandeducationbecausetheyaredemographic/socioeconomicvariablesthatarenotavailableinadministrativerecordsandareimportantvariablesthattheCensusBureaucollectsdataon.
DRAFT
24
purposesonlyandhavenotundergoneformalstatisticaltesting.Fortransparency,weuse
unweighteddata.Otheranalystsmaydisagreewithourassessmentofthevisualevidence.
Figures3through6showdifferencesinoursamplebyeducationalattainment,starting
withoverallparticipationratesandthenexploringdifferencesincoverage,take-upandaverage
contributions.Figure3showsthatforeverygroup,participationratesincreaseacrossincome
deciles.Thedifferencebetweenparticipationratesbyeducationalattainmentappearlargerat
highincomedecilesrelativetolowerincomedeciles.Turningtoofferrates,Figure4shows
workerswithoutaHSdiplomaarelesslikelytoworkatfirmsthatofferDCplansthanare
workerswithaHSdiploma,acrosstheincomedistribution.Also,foralleducationlevels,lower
incomeworkersarelesslikelytobeatfirmsofferingDCthanarehigherincomeworkers.
Next,welookattake-up,whichisparticipationconditionalonbeingofferedaplan.
Figure5showsthatforthelowestdecilesofincome,take-upislowregardlessofeducation.At
higherincomedeciles,take-upratesdivergebyeducation.However,take-upratesarehigher
forcollegegraduatesthanthosewithgraduatedegrees,particularlyatthemiddleincome
deciles.ThiscouldreflectthefactthatwedonotobserveRothcontributionsandthereforewe
underestimatetakeupamonggroupswithRothcontributions(andnobefore-taxcontributions).
Taxpayerswithtemporarilylowincomehaveincentivetocontributeafter-taxratherthan
before-taxincome,andthiscouldbethecaseforgraduatedegreeholdersinthemiddledeciles
ratherthanthetopdecilesofincome.
Figure6showstheaverageshareofearningscontributedtoaplan,conditionalon
contributing.Workersarestratifiedbytotalearningsatalljobs,whereascontributionratesare
calculatedusingdeferredcompensationandtotalearningsatthemainjob.Amongworkers
withtax-deferredcontributions,thecontributionlevelishigherforcollegeandgraduatedegree
holders,acrossincomedeciles.Thelargestdifferencesincontributionratesareatthelowest
incomedeciles.
Nextweturntowithin-incomebindifferencesbyrace,presentedinFigures7through
10.Lookingfirstatparticipationrates(Figure7),thetrajectoryofparticipationratesacross
incomeissimilaracrossrace.Thepatternsforofferrates(Figure8)aresurprising.Black
DRAFT
25
workersaremorelikelytobeofferedDCplans,acrosstheincomedistribution.Atlowincome
deciles,Asianworkersarelesslikelytobeofferedplans,butthenofferratesforAsiansare
indistinguishablefromothergroupsathigherincomedeciles.Forallraces,lowerincome
workersarelesslikelytobeatfirmsofferingDCthanarehigherincomeworkers.
Figure9showsthattake-upratesarelowforallgroupsinthebottomincomedeciles,
andallgroupsthereisincreasingtake-upacrossincomedeciles.Acrossincomedecilesthereis
highertake-upamongwhitesandAsians.Amongworkerswithtax-deferredcontributions,
Figure10showsthecontributionlevelishigherforAsianandwhite(non-Hispanic)workers,
acrossincomedeciles.Largestdifferencesacrossrace/ethnicityareinthelowestincome
deciles.
TheselargedifferencesinofferratesbyracecouldbeinfluencedbyBlackworkerson
averagebeingemployedatlargerfirms.InFigure11,weplotofferratesagain,thistime
controllingforfirmsize.Theresidualizedincomesarebinnedandplotted.ThebinsareNOT
necessarilythesameasintheFigureabove,sotheexactlevelsarenotdirectlycomparable.
Nevertheless,thehigherofferratesforBlackworkersdisappearaftercontrollingforfirmsize,
asdothelowerofferratesforAsians(inlowincomedeciles).Thesechangeswhencontrolling
forfirmsizestemfromdifferencesincoverageacrossfirmsize,asillustratedearlier,andthe
compositionofthefirmsemployingworkersatdifferentincomelevelsandracegroups.
6. ConclusionInthispaper,wepresentestimatesofDCplansponsorshipandparticipationusingthe
universeofW-2records.Toourknowledge,thisisthefirsttimetaxdatahavebeenusedfor
thispurpose.OurmaincontributionisdevelopingnewmeasuresofDCplansponsorshipthat
generatebetterestimatesofDCcoverageandparticipation.
OurestimatesofDCplancoverageandtake-updivergefromestimatesusingotherdata.
UsingW-2records,weestimatefirmsponsorshipratesthataresmallerthanhasbeenfound
usingemployersurveys(NCS,2012).Atthesametime,thejobsandindividuallevelcoverage
ratesarehigherthanhasbeenfoundelsewhere(Dushietal.,2011).Weshowthatindividuals
DRAFT
26
intheCPSunderreportretirementplancoveragerates,whichleadstoinflatedtake-uprate
estimates.Theseresultshaveimportantimplicationsforunderstandingtake-upand
heterogeneityacrossdemographicandsocioeconomicgroups.
Aswehaveshown,employer-employeelinkeddataprovideimportantinformation
neededtobetterunderstandsponsorship,participationandthepotentialimpactofpolicies
incentivizingfirmsponsorship.Estimatesofnationwidepensioncoveragevarysubstantially
dependingonthesourceofdata,particularlywhethertheunitofanalysisistheworkerorthe
firm.OurestimatesofDCsponsorshipamongsmallfirmsdivergefromtheNational
CompensationSurvey(NCS)estimates,highlightingpotentialstrengthsofusingtheuniverseof
firmsorplansratherthananationallyrepresentativefirm-levelsurvey.Withsmallfirmsatthe
heartofpolicydiscussions,additionalworkinvestigatingthesedifferencesiscriticalto
producingbetterestimatesofthepotentialeffectsofpolicychanges.
DRAFT
27
References1. Aaronson,Stephanie,andJuliaLynnCoronado.2005.“AreFirmsorWorkersBehindthe
ShiftAwayfromDBPensionPlans?”FederalReserveBoardFinanceandEconomics
DiscussionSeriesPaper2005-17.
2. Allen,Steven,RobertClark,andAnnMcDermed.1993.“PensionBondingandLifetime
Jobs.”JournalofHumanResources28(3):463-481.
3. Chetty,Raj,JohnN.Friedman,SørenLeth-Petersen,TorbenHeienNielsen,andToreOlsen.
2014.“Activevs.PassiveDecisionsandCrowd-OutinRetirementSavingsAccounts:
EvidencefromDenmark.”QuarterlyJournalofEconomics129(3):1141–1219.
4. Choi,JamesJ.2015.“ContributionstoDefinedContributionPensionPlans.”AnnualReview
ofFinancialEconomics7(1):161–78.
5. Choi,JamesJ.,DavidLaibson,BrigitteC.Madrian,andAndrewMetrick.2004.“ForBetteror
forWorse:DefaultEffectsand401(k) SavingsBehavior.”InperspectivesontheEconomicsof
Aging,editedbyDavidA.Wise,81–126.Chicago:UniversityofChicagoPress.
6. Copeland,Craig.2015.“TheEffectoftheCurrentPopulationSurveyRedesignon
Retirement-PlanParticipationEstimates.”EBRINotes,Vol.36,No.12.
7. Dummann,Kathrin.2008.“Whatdeterminessupplyanddemandforoccupationalpensions
inGermany?”JournalofPensionEconomicsandFinance7(2):131–56.
8. Dushi,IrenaandHowardM.Iams.2010.“TheImpactofResponseErroronParticipation
RatesandContributionstoDefinedContributionPensionPlans.”SocialSecurityBulletin,
Vol.70No.1.
9. Dushi,Irena,HowardM.Iams,andJulesLichtenstein.2011.“AssessmentofRetirementPlan
CoveragebyFirmSize,UsingW-2TaxRecords.”SocialSecurityBulletin,Vol.71,No.2.
10. Fadlon,Itzik,JessicaLaird,andTorbenHeienNielsen.2016.“DoEmployerPension
ContributionsReflectEmployeePreferences?EvidencefromaRetirementSavingsReformin
Denmark”AmericanEconomicJournal:AppliedEconomics.8(3):196-216.
11. Gustman,AlanL.,ThomasL.Steinmeier,andNahidTabatabai.2009.“DoWorkersKnow
aboutTheirPensionPlanType?ComparingWorkers’andEmployers’PensionInformation.”
Inovercomingthesavingsslump:HowtoIncreasetheEffectivenessofFinancialEducation
DRAFT
28
andsavingprograms,editedbyAnnamariaLusardi,47–81.Chicago:UniversityofChicago
Press.
12. Ippolito,Richard.1991.“EncouragingLong-TermTenure:WageTiltorPensions?”Industrial
andLaborRelationsReview44(3):520-535.
13. Jarmin,RonandJavierMiranda.2002.“TheLongitudinalBusinessDatabase.”
https://www.census.gov/ces/pdf/CES-WP-02-17.pdf.
14. Madrian,BrigitteC.andDennisF.Shea.2001.“ThePowerofSuggestion:Inertiain401(K)
ParticipationandSavingsBehavior.”TheQuarterlyJournalofEconomics.116(4):1149-1187.
15. Meyer,BruceD.,WallaceK.C.Mok,andJamesX.Sullivan,2015.“HouseholdSurveysin
Crisis.”JournalofEconomicPerspectives.29(4):199-226.
16. Mitchell,OliviaS.1988.“WorkerKnowledgeofPensionProvisions.”JournalofLabor
Economics6(1):21–39.
17. Nessmith,WilliamE.,StephenP.Utkus,andJeanA.Young.2007.“Measuringthe
EffectivenessofAutomaticEnrollment.”Volume31.ValleyForge,PA:VanguardCenterfor
RetirementResearch.
18. NationalCompensationSurvey:EmployeeBenefitsintheUnitedStates,March2012.
Bulletin2773.
19. Papke,LeslieE.1995.”ParticipationinandContributionsto401(k)PensionPlans:Evidence
fromPlanData,”TheJournalofHumanResources,Vol.30,No.2(Spring1995),pp.311-325.
20. Song,Jae,DavidPrice,FatihGuvenen,andNicholasBloom.2015.“FirmingupInequality.”
NBERworkingpaper.
21. TheWhiteHouse,OfficeofthePressSecretary.“FACTSHEET:Buildinga21stCentury
RetirementSystem.”ReleasedJanuary26,2016.http://www.whitehouse.gov/the-press-
office/2016/01/26/fact-sheet-building-21st-century-retirement-system
22. VanDerhei,Jack.2010.“TheImpactofAutomaticEnrollmentin401(k)PlansonFuture
RetirementAccumulations:ASimulationStudyBasedonPlanDesignModificationsofLarge
PlanSponsors.”IssueBrief341.Washington,DC:EmployeeBenefitResearchInstitute.
23. Wagner,DeborahandMaryLayne.“ThePersonIdentificationValidationSystem(PVS):
ApplyingtheCenterforAdministrativeRecordsResearchandApplications’(CARRA)Record
DRAFT
29
LinkageSoftware.”CARRAWorkingPaperSeries.WorkingPaper#2014-01.ReportIssued:
July1,2014.
DRAFT
30
TablesTable1:PopulationcountsandDCplancoverageinW-2records(2012)
Firms(EINs) NumberinW-2records 6.2MPercentofEINsofferingDCplans 13.7
Jobs NumberinW-2records 218.9MPercentofjobsatEINwithDC 71.2Take-upamongjobsatEINswithDC 33.6
Individuals NumberinW-2records 154.8MPercentofworkersatEINswithDC 78.5Take-upamongworkersatEINswithDC 41.4
Individualswithover$3,770totalearnings NumberinW-2records 135.1MPercentofworkersatEINswithDC 80.8Take-upamongworkersatEINswithDC 45.7
NOTE:AjobisatanEINthatsponsorsaDCplaniftherearedeferredearningsfromatleastonejobattheEIN.AworkerisofferedaplanifemployedatanyfirmofferingDCplansandparticipatesiftheyhavedeferredcompensationatanyoftheirjobs.SOURCE:U.S.CensusBureau,TY2012FormW-2
Table2:DCCoverageandparticipation,byfirmsize(2012W-2s)
PanelA:EINs-level PanelB:Jobslevel
EINsize(numberofemployees)
PercentofEINs Offered
Averagetake-up
Percentofjobs Coverage
1 23.61 3.4 100.0 0.67 3.42–9 46.91 7.3 55.2 5.72 8.910–24 16.04 20.2 41.1 6.89 21.125–49 6.45 32.6 36.9 6.32 33.350–99 3.48 46.5 33.6 6.82 47.4100–499 2.84 68.5 32.5 16.13 72.1500-1,000 0.35 83.5 34.7 6.82 83.81,000-10,000 0.3 88.4 35.1 21.23 89.710,000+ 0.03 92.9 32.3 29.4 93.5TotalCount 6.2M 13.7 45.1 218.9M 71.2NOTE:Firmsaregroupedbythenumberofemployees,whichisdeterminedfromthenumberofuniquePIKsintheW2recordsassociatedwiththeEINin2012.SOURCE:U.S.CensusBureau,TY2012FormW-2
DRAFT
31
Table3:Jobs-levelcoverageandparticipationacrossearnings
Earningscategories
AtfirmofferingDCplan(percent)
ParticipatinginDCplan(percent)
Take-uprate(percent)
$1-10k 63.6 5.6 8.8$10-20k 65.1 15.9 24.4$20-30k 75.2 31.0 41.2$30-50k 82.8 45.1 54.5$50-75k 89.3 59.3 66.4$75-100k 92.0 69.4 75.4$100-150k 92.7 76.4 82.4$150-200k 91.9 78.2 85.1$200-300k 90.4 77.2 85.4$300-500k 90.8 76.9 84.7$500k+ 91.2 75.3 82.5NOTE:Thesenumbersarebasedonjob-levelannualearningsfromthepopulationofW-2recordsin2012.Jobsareplacedintoearningscategories.Individualsmaythereforeshowupmorethanonce,butfordifferentfirms.SOURCE:U.S.CensusBureau,TY2012FormW-2Table4:PercentofEINswithDCplansforuntenuredemployees,byfirmsizeEINsize(numberofemployees) AllEmployees 1st-YearEmployeesTotal 27.8 9.3Fewerthan10 8.2 1.110–24 20.0 3.225-49 34.4 8.550-99 49.4 17.6100ormore 74.2 45.2NOTE:ThesampleisrestrictedtoEINsthatexistfor2010,2011and2012,withatleastonefirst-yearemployee(in2012only)andatleastonewhowasnot.SOURCE:U.S.CensusBureau,TY2012FormW-2
DRAFT
32
Table5:HeterogeneityinDCPlanSponsorshipandParticipationRatesacrossEINs(in2012)
PercentoffirmsofferingDC
plans
AvgParticipationRateacrossfirmsofferingplans
PanelA:Typeoffirm SoleProprietors 3.3 47.7
S-Corporation 12.7 48.8Partnership 14.2 41.3C-Corporation 19.2 48.8Tax-exempt 19.4 46.1Non-profit 28.3 30.8State/LocalGov't 51.1 29.3
PanelB:ByNAICSsector Agr,Forest,Fish 3.0 41.0Accommodation&Food 3.6 13.0RetailTrade 8.3 34.7Construction 9.5 36.8RealEstate 9.8 51.6Arts,Entertainment 10.5 32.0Transportation 10.7 40.1Administration 10.9 41.1OtherServices 11.5 34.7Professional 18.4 59.3Mining 19.4 55.8Finance&Insurance 20.5 61.2WholesaleTrade 20.7 52.0HealthCare 23.1 43.5Information 24.1 51.7Manufacturing 26.5 46.9EducationServices 30.6 26.6Utilities 36.7 61.1Management 39.3 62.1PublicAdmin 40.4 32.9
NOTE:Firmsaregroupedbythenumberofemployees,whichisdeterminedfromthenumberofuniquePIKsintheW2recordsassociatedwiththeEINin2012.Categoriesfornon-profit,federalgovt,state/localgovernmentaredeterminedbytheEmployerCode,whilecategorizationsforthedetailedcategoriesarefromLegalFormofOrganization(LFO).EINsassociatedwiththefederalgovernmenthavebeenremoved,accordingtoCensuspolicy.FortheNAICSsector,thecategoryfor"other"isnotreportedduetosmallcellsize.Theaverageparticipationisatfirmsthatofferplans.SOURCE:U.S.CensusBureau,EIN-linkedTY2012FormW-2and2012BusinessRegister
DRAFT
33
Table6:PensioncoverageusingW-2andForm5500(2012)
UsingEIN-levelmeasures
UsingFirm-levelmeasures
EINs Jobs EINs JobsNone 84.4 31.6 84.2 30.1DC(W2) 6.5 30.2 5.5 11.4DC(5500) 1.6 1.5 1.6 1.5DC(W2,5500) 7.0 28.5 7.8 38.1DB(5500) 0.1 0.1 0.1 0.1DC(W2);DB(5500) 0.0 0.7 0.0 0.2DC(5500);DB(5500) 0.1 0.1 0.1 0.1DC(W2,5500);DB(5500) 0.3 7.4 0.7 18.5TOTALOBS 5.9M 182.5M 5.9M 182.5MNOTES:PublicsectorjobsareexcludedfromthesamplebecausetheyareexemptfromForm5500filingrequirements.SOURCE:U.S.CensusBureau,EIN-linkedTY2012FormW-2and2012Form5500ResearchFile
Table7:AutoenrollmentinDCplanscoverage(2012)
UsingEIN-level
measuresUsingFirm-level
measuresDCcoveragecategorization(percent) EINs Jobs EINs JobsDCcoverage-autoenrollment 0.4 6.3 0.7 13.4DCcoverage-noautoenrollment 6.9 29.5 7.6 41.8DCcoverage-unknownautoenrollment 6.5 30.9 5.5 11.5NoDCcoverage 86.2 33.2 86.2 33.2Totalcount 5.9M 182.5M 5.9M 182.5MNOTE:PublicsectorjobsareexcludedfromthesamplebecausetheyareexemptfromForm5500filingrequirements.WeuseForm5500informationassociatedwiththeEIN.IftherearenoactiveDCplansassociatedwiththeEIN,weappendinforrnationontheactiveDCplansassociatedwithotherEINsassociatedwiththesamefirm.SOURCE:U.S.CensusBureau,EIN-linkedTY2012FormW-2and2012Form5500ResearchFile
Table8:Agreementof2013CPS-ASECandW-2RetirementPlanInformationforPrivateSectorWorkersAge25-64
Coverage Participation
All Non-Imputed Imputed All Non-
Imputed Imputed
W-2:yes;CPS:yes 0.469 0.481 0.42 0.277 0.297 0.195W-2:yes;CPS:no 0.277 0.27 0.306 0.116 0.105 0.158W-2:no;CPS:yes 0.057 0.042 0.118 0.158 0.137 0.248W-2:no;CPS:no 0.196 0.206 0.156 0.449 0.461 0.399SOURCE:U.S.CensusBureau,PIK-linked2013CPSASECpersonrecordsandTY2012FormW-2
DRAFT
34
Table9:Comparisonof2013CPS-ASECandW-2RetirementPlanInformationforWageandSalaryWorkersAge25-64byDemographicGroup
OfferRate Participation
Rate Take-UpRate
CPS W-2
CPS W-2
CPS W-2
PublicSector 0.81 0.94
0.74 0.43
0.91 0.46PrivateSector 0.53 0.75
0.44 0.39
0.83 0.53
PrivateSubgroups
Age25to34 0.47 0.75
0.35 0.32
0.74 0.43Age35to44 0.53 0.74
0.45 0.41
0.84 0.55
Age45to54 0.56 0.75
0.49 0.43
0.87 0.57Age55to64 0.56 0.75
0.48 0.44
0.86 0.58
<HSGrad 0.26 0.55
0.17 0.16
0.67 0.29HSGrad 0.46 0.70
0.36 0.31
0.79 0.45
SomeCollege 0.54 0.76
0.44 0.39
0.81 0.51CollegeGrad 0.64 0.82
0.56 0.53
0.88 0.64
White 0.54 0.74
0.45 0.40
0.84 0.54WhiteNon-Hispanic 0.58 0.77
0.49 0.43
0.84 0.57
Black 0.48 0.79
0.37 0.34
0.76 0.43Asian 0.50 0.70
0.43 0.41
0.86 0.59
HispanicAnyRace 0.35 0.64
0.27 0.25
0.77 0.40
Full-timeFull-Year 0.59 0.79
0.52 0.47
0.87 0.60NotFull-timeFull-Year 0.35 0.64
0.21 0.19
0.62 0.30
FirmSize1-10 0.18 0.36
0.15 0.12
0.79 0.35Firmsize10-24 0.34 0.59
0.27 0.24
0.80 0.42
FirmSize25-99 0.48 0.74
0.39 0.33
0.82 0.45FirmSize100-499 0.58 0.84
0.48 0.43
0.82 0.51
FirmSize500-999 0.63 0.89
0.53 0.48
0.84 0.54FirmSize1000+ 0.71 0.90
0.59 0.54
0.84 0.60
Union 0.74 0.90
0.66 0.43
0.90 0.47NotUnion 0.52 0.75 0.43 0.42 0.82 0.56SOURCE:U.S.CensusBureau,PIK-linked2013CPSASECpersonrecordsandTY2012FormW-2
DRAFT
35
Table10:IndividualcoverageandparticipationbyACSdemographiccharacteristics
Shareof
sample Coverage ParticipationTake-up
Contributionrate
All
75.4 43.4 57.7 6.9Education
NoHSdiploma 6.8 61.4 22.9 37.3 5.2HSdiploma/GED 26.3 71.5 36.6 51.2 5.9Somecollege 33.4 75.5 42.7 56.6 6.3Collegegraduate 22.4 80.5 53.4 66.3 7.6Graduatedegree 11.1 82.3 54.3 66.0 8.7
Race/ethnicity White,non-
hispanic 75.1 75.6 45.6 60.4 7.0Black 8.4 80.8 36.7 45.5 5.1Asian 5.4 74.1 46.6 62.9 8.9Hispanic 9.8 70.2 32.1 45.6 5.5Other 1.2 71.5 32.9 46.1 5.9
NOTE:Thereareroughly1,320,000observationsfromACS2012.Theshareofpopulationistheunweightedcountoverthetotalnumberofobservations.SOURCE:U.S.CensusBureau,PIK-linked2012ACSpersonrecordsandTY2012FormW-2
Table11:Comparingestimatestopreviouswork PanelA:Firms'offerratesusingW-2sandNCS(inpercent) Allfirms 1-99employees 100+employeesW-2s 14 12* 72*W-2s–restricted** 28
74
NCS(2012) 43 42 74PanelB:Workersofferandparticipationrates
Offered Participation W-2s:job-level 71 24
W-2s:PIK-level^ 79 32W-2s:PIK-level,restricted^^ 81 37NCS(2012) 55 37SIPP(2006) 57 39
SOURCE:SIPP(2006)fromDushietal.(2011),Table2:SIPPrespondentsage21-64withmainjoborbusinessin2006.NationalCompensationSurvey(NCS),2012:allworkers.^workerisofferedaplanifemployedatanyfirmofferingDC;participatesifDCcontributionatanyjob.**Restrictedtofirmsin2010-2012withatleastonetenuredandone1st-yearemployee.^^Excludesworkerswith<$3,770totalearnings.
DRAFT
36
FiguresFigure1:Firmsponsorshipandparticipationrates:ByFirmSize&VentilesofMeanEarnings(byEINSize)(A.)PercentofEINsofferingDCplans
(B.)AverageShareofWorkersParticipating(amongFirmsOfferingDCPlans):
SOURCE:U.S.CensusBureau,TY2012FormW-2
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Venblesofmeanearnings(byEINsize)
100ormore
Fewerthan10
10–99
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Venblesofmeanearnings(byEINsize)
100ormore
Fewerthan10
10–99DRAFT
37
Figure 2: CPS-ASEC and W-2 Pension Plan Information Over Time (A.)Coverage
(B.)Participation
(C.)Take-Up
NOTE:Samplerestrictedtoprivatesectorworkersage25-64SOURCE:U.S.CensusBureau,PIK-linked2013CPSASECpersonrecordsandTY2012FormW-2
0.570.53
0.76 0.75
2005 2006 2007 2008 2009 2010 2011 2012
CPS
W-2
0.47
0.44
0.370.39
2005 2006 2007 2008 2009 2010 2011 2012
CPS
W-2
0.82 0.83
0.490.53
2005 2006 2007 2008 2009 2010 2011 2012
CPS
W-2
DRAFT
38
Figure3:Participationratesacrossearningsandeducation
Figure4:Coverageratesacrossearningsandeducation
DRAFT
39
Figure5:Takeupratesacrossearningsandeducation
Figure6:Averagecontributionamongparticipantsbyearningsandeducation
DRAFT
40
Figure7:Participationratesacrossearningsandrace
Figure8:Coverageratesacrossearningsandrace
DRAFT
41
Figure9:Takeupratesacrossearningsandeducation
Figure10:Averagecontributionamongparticipantsbyearningsandrace
DRAFT
42
Figure11:Coverageratesacrossearningsandrace,conditionalonfirmsize
DRAFT
top related