Вы находитесь на странице: 1из 53

THE LABOR FORCE

PARTICIPATION RATE
SINCE 2007:
CAUSES AND POLICY IMPLICATIONS


July 2014



Embargoed for 10 am on July 17
th
, 2014
2

Introduction and Executive Summary



In 2008, the U.S. economy collided with two historic forces. The first force was the Great
Recession,themostsevereeconomiccrisisinageneration.Whiletheeconomyhasrecovered
considerablyoverthelastfiveyears,thereislittledoubtthatmoreworkremainstoaddresssome
of the challenges left in the wake of the Great Recession. The turmoil of 2008 inflicted
tremendouspainonmillionsoffamilies,overshadowingthefactthat2008alsomarkedaunique
milestoneinU.S.economichistory.Thatyear,thefirstbabyboomers(thosebornin1946)turned
62 and became eligible for Social Security early retirement benefits. This second forcethe
demographicinflectionpointstemmingfromtheretirementofthebabyboomerswasfeltfar
less acutely than the Great Recession, but will continue to have a profound influence on the
economyforyearstocome,wellafterthebusinesscyclerecoveryfromtheGreatRecessionis
consideredcomplete.

Inadditiontotheseinflectionpointsin2008,anumberoflongertermtrendshadbeenplaying
outintheU.S.laborforcepriorto2008andhavecontinuedsincethen.Theseincludethenearly
continuousdeclineinlaborforceparticipationratesforprimeagemales(i.e.,age2554)since
themid1950sandthedramaticriseinlaborforceparticipationratesforprimeagefemalesin
the1970sand1980sfollowedbyastallingandslighttrenddeclineafterthelate1990s.

Manydimensionsoftheeconomysperformanceoverthelastseveralyearscanonlybeproperly
evaluatedwhentheeffectsoftheGreatRecession,theretirementboom,andthelongerterm
laborforcetrendsaretakenintoaccount.Oneoftheclearestillustrationsofthispointisthelabor
forceparticipationrate,whichrepresentsthefractionoftheadultpopulationeitherworkingor
lookingforwork.Changesinlaborforceparticipationreflectnotjustcurrenteconomicconditions
likejobavailabilityandworkersassessmentsofjobfindingprospects,butalsomorestructural
factors like the age distribution of the population and other aspects of society that impact
peoplesdecisionstoparticipateinthelaborforce.

This report analyzes the evolution of the labor force participation rate since late 2007 and
attemptstoquantifytheeffectsofthesevariousforces.Weexaminetheperiodsince2007to
focusonhoweachofthetwolargestforces,theGreatRecessionandtheretirementofthebaby
boomers,hasimpactedlaborforceparticipationinrecentyears.Wefindthatthecombination
ofdemographicchangesandthedropinlaborforceparticipationthatwouldhavebeenexpected
basedonhistoricalbusinesscyclepatternsexplainmostbutnotalloftherecentdropinlabor
force participation. This implies that other factors, likely including both a continuation of pre
existingtrendsinlaborforceparticipationbycertaingroupsandtheuniqueeffectsoftheGreat
Recessionhavealsobeenimportant.Thisreportalsodiscussesthelaborforceparticipationrates
fordifferentgroups,discussespotentialfuturescenariosfortheparticipationrate,andlaysout
policiesthatwouldhelptoboostparticipationintheyearstocome.

Summary of the Main Results


Sincethefinalquarterof2007,thelaborforceparticipationratehasfallenfrom65.9percentto
62.8percentinthesecondquarterof2014,adeclineof3.1percentagepoints.Inthisreport,the
CouncilofEconomicAdvisersestimatesthatthis3.1percentagepointdeclinecanbeattributed
tothreemainsources:

Abouthalfofthedecline(1.6percentagepoint)isduetotheagingofthepopulation.

Because older individuals participate in the labor force at lower rates than younger
workers,theagingofthepopulationexertsdownwardpressureontheoveralllaborforce
participationrate.Whileolderworkerstodayareparticipatinginthelaborforceathigher
rates than older workers of previous generations, there is still a very large dropoff in
participationwhenworkersentertheirearly60s.

About a sixth of the decline in the overall participation rate (0.5 percentage point) is a cyclical
declineinlinewithhistoricalpatternsinpreviousrecessions.

While the unemployment rate has come down from a peak of 10.0 percent in October
2009 to 6.1 percent in June 2014, it has remained elevated for the last several years.
Historically,elevationintheunemploymentrateisassociatedwithadeclineinlaborforce
participation,aspotentialworkersmaydecidetodeferlookingforajobuntiltheeconomy
improves. A portion of the most recent decline in the participation rate reflects this
historicalpattern,whichthisreportreferstoasthecyclicaleffect.

Aboutathirdofthedecline(1.0percentagepoint)arisesfromotherfactors,whichmayinclude
trendsthatpredatetheGreatRecessionandconsequencesoftheuniqueseverityoftheGreat
Recession.

Inparticularthetwoelementsofthisresidualnotexplainedbythestandardfactorsare:

Thefactthatparticipationratesconditionalonageweredecliningformanygroupsin
therunupto2008,includingforprimeagemenfromthe1950sandforprimeage
womenfromthelate1990s,mayalsohavecontributedtothedeclineinparticipation.
Thiswouldhavebeenexpectedtoresultinadeclineintheparticipationrateabove
and beyond the pure aging effect even in the absence of a recession. Note, these
effectswerepartlyoffsetbyotherpreexistingtrends,likeariseintheparticipation
rateforolderworkers.

TheseverityoftheGreatRecession,whichhasresultedinanunusualpatternofavery
large share of longterm unemployed relative to total unemployment, may have
loweredtheparticipationratemorethanwouldhavebeenexpectedthroughnormal
channels.Infact,wefindthatadecliningparticipationrateishistoricallycorrelated
with the elevation of longterm unemployment, suggesting that both issues have a
commoncauseorcauseeachother.
4

Inthelastfewyearstheagingofthepopulationhasbeenanincreasinglyimportantsourceof
thedeclineintheparticipationrate.Fromthebeginningof2011tothesecondquarterof2014,
theparticipationratefellby1.4percentagepoints.Around70percentofthatdecline(1.0
percentagepoint)canbedirectlyattributedtotheagingofthepopulationandincreased
retirements.

Overview of the Report


SectionIofthisreportprovideshistoricalcontextwithabroadlookattheparticipationratesince
1948.ThenSectionIIsurveystheliteratureontheparticipationrateintheGreatRecessionand
subsequentrecovery.Mostresearchershavefoundthatlongtermtrendsincludingtheagingof
thepopulationexplainamajorityoftherecentdeclineinparticipation,andthatcyclicaleffects
explainaboutaquarterofthedecline.SectionIIIexplainsCEAsdecompositionofmovementsin
theparticipationrateintotheagingofthepopulation,thebusinesscycle,andotherfactors.

Section IV takes a deeper look at the trends in participation for several demographic groups:
young men and women, older workers, primeage women, primeage men, and black men.
Participationamongyoungmenandwomenhasdecreasedsteadilyoverthepastdecade,largely
reflectingdecreasedparticipationamongstudentsandincreasedenrollment.Meanwhile,older
workers have been participating at increasing rates over the last few decades. Black male
participation has been on a troubling downward trend since the mid1970s, falling about 10
percentage points over that time, while a similar, albeit slightly more muted decline has been
seenamongallmen.

SectionVlaysoutpotentialscenariosfortheparticipationrategoingforwardassumingnopolicy
changesaremade.Theagingofthepopulationwilllikelycontinuetoexertdownwardpressure
ontheoverallparticipationrate.Atthesametime,continuedrecoveryoftheeconomyfromthe
GreatRecessionwouldputupwardpressureonparticipation.Assumingnochangeinpolicywe
estimatethattherecoveryintheparticipationratefromcyclicalfactorswillberoughlyoffsetby
theagingeffect,leadingtoamostlyunchangedparticipationrateonnet.However,itisdifficult
toknowhowmuchrecoveryintheresidualwilloccurastheeconomyimproves,andparticipation
may rise as those additional workers are pulled back into the labor force. Once the cyclical
reboundfades,however,assumingnopolicyorotherstructuralchanges,theparticipationrate
will resume its downward trend driven mainly by aging. This discussion lays out a baseline for
whatmighthappenintheabsenceofpolicystepstopromoteparticipationintheworkforce.

Section VI lays out policy measures that can boost participation and in so doing, expand the
economyslongrungrowthpotential.First,immigrationreformwouldraisethesizeofthelabor
forceaswellasboostparticipationratesandcouldlargelyoffsetfurtherdeclinesinparticipation
duetoaging.Second,adoptingfamilyfriendlyworkplacepoliciescanboostfemalelaborforce
participation. For instance, many states are currently considering adopting paid leave policies,
andevidencesuggeststhattheavailabilityofpaidleaveandotherfamilyfriendlypoliciescould
increasetheparticipationofwomenbyabout7percentagepoints(BlauandKahn,2013).Third,
thelongrundeclineinparticipationamongmen,especiallyminoritymen,isunsustainable,and
5

the Obama Administration is aggressively pursuing policies aimed at stemming these declines.
Finally,inthenearterm,therecoveryfromtheGreatRecessionremainsincomplete,andsteps
tostrengthentheeconomyandencouragebestpracticesforhiringthelongtermunemployed
canstillmakeamajordifference.Thesetypesofstepsarecriticalforcombatingtheportionof
thedeclineintheparticipationratethatreflectstheuniqueseverityoftheGreatRecessionand
themajorchallengesleftinitswake.

I. The Labor Force Participation Rate Over the Past 50 Years

Thelaborforceparticipationrateistheshareofthecivilianpopulationoverage15thatiseither
working or searching for work (i.e., unemployed). Increases in the participation rate reflect
individualsjoiningtheworkforce,eitherbylookingforajobortakingone,whiledecreasesinthe
participationratecanreflectanincreaseinretirements,peopletakingtimeoutofthelaborforce
for a variety of reasons (e.g., to go to school, care for a young child, etc.), or unemployed
individuals who become discouraged and stop looking for a job. Decreased labor force
participation may or may not be desirable, depending on the reason. Achieving a labor force
participationrateof100percentisnotthegoal,sincemanypeoplewholeavethelaborforceare
engaged in important nonmarket activities, such as schooling, raising children, or retirement.
However,discouragedworkersdroppingoutofthelaborforcewouldbeacauseforconcernand
wouldmotivateenactingpoliciestoreturntheseworkerstothelaborforce.Inthelongrun,the
growth rate of the labor force underpins the growth rate of employment, which, along with
productivitygrowth,isakeydeterminantofthegrowthrateofpotentialGDP.

ThelaborforceparticipationrateintheUnitedStatessince1948hasthreekeyfeatures,asshown
inFigure1.ThefirstisthestrikinginvertedUshapedtrendintheparticipationrate:whilejust
under59percentofcivilianadultswereinthelaborforcein1963,thisrosetoapproximately66
percent in 1990 and peaked at 67 percent in 2000, but since 2000 the participation rate has
generallyfallen.Asdiscussedinmoredetailbelow,thisinvertedUshapedpatterncanbemore
fullyunderstoodbylookingattheseparatetrendsformenandwomen.Ingeneral,theleft(rising)
sideoftheinvertedUshapereflectsahistoricsurgeofwomenintothelaborforceinthesecond
halfofthetwentiethcentury,aswellasthebabyboomersreachingtheirprimeworkingyears.
The right (falling) side reflects the plateauing of female participation, combined with the
retirementofthebabyboomgenerationandthemassiveshockoftheGreatRecession.

58
60
62
64
66
68
70
1948 1958 1968 1978 1988 1998 2008
Figure1:TotalCivilianLaborForceParticipationRate
Percent
2014:Q2
7

ThesecondkeyfeatureoftheparticipationratevisibleinFigure1istheprocyclicalpattern.In
otherwords,participationrisesduringanexpansionandfallsduringa recession.Forexample,
thesecyclicalpatterns,whichoccurwithalag,areclearlyvisibleinthe1990recession.During
thestronglaborforcegrowthofthe1970sand1980s,theprocyclicalmovementsdonotproduce
distinctdeclinesintheparticipationrate,butinsteadproducetemporaryplateaus,asthecyclical
declinewasmitigatedbyanupwardunderlyingtrendthatresumedaftertherecoverywasunder
way.Similarly,becausetheparticipationratewasonadownwardtrajectoryduringthefirsthalf
of the 2000s, the cyclical decline during and after the 2007 recession combined with the
underlyingdemographictrendtoproduceaparticularlysharpdrop.

ThethirdfeatureoftheparticipationratevisibleinFigure1isthattheseriesexhibitsconsiderable
volatilityaroundthelongtermtrendsandcyclicalmovements.Forexample,duringtherecovery
fromthe1990recession,theparticipationraterosemorethanhalfapercentagepointinthefirst
halfof1992,onlytosubsequentlyfallbynearlythesameamount.Muchofthisvolatilitystems
from factors like the normal statistical noise that exists in any data series based on a limited
sampleandseasonallyadjusted.

As noted earlier, the inverted Ushaped trend in the overall participation rate can be better
understoodbyconsideringtheseparatetrendsformaleandfemalelaborforceparticipation.
Themalelaborforceparticipationratehasbeenfallingsincedatacollectionbeganin1948,as
showninFigure2.Someofthisdeclinereflectstheincreasinglongevityofmenovertimeandthe
resulting lengthening of their retirement. However, the rate of participation among primeage
menhasalsofallenoverthistimeperiod,althoughtoalesserextent.

In contrast, womens labor force participation rose through much of that period, with a surge
from the mid1960s through the 1990s. This profound shift in Americas economy and society
reflected changes in social norms and access to education, including the legalization of oral
contraceptivesintheearly1960s,reductionsindiscriminationthroughlegislationlikeTitleVIIof
the Equal Rights Act and Title IX of the Educational Amendments to the Equal Rights Act, and
developments in laborsaving household technology. The rise in womens labor force
participation exceeds the decline in mens participation so that overall adult participation
increasedduringthisinterval.
8

Since2000,theparticipationratesofmenandwomenhavebothdeclined,althoughtherateof
declinehasbeenslowerforwomenthanformen.Participationrateshavelongbeeninfluenced
byanagingpopulation,howeveritsimpactbecamemorepronouncedstartingin2008whenthe
firstcohortofbabyboomers(thosebornin1946)turned62andbecameeligibleforearlySocial
Securitybenefits.Startingin2008,theshareoftheworkingagepopulationoverage55beganto
rise rapidly (Figure 3). Also, while the share of the population over age 65 has risen gradually
since1960duetoincreasedlongevity,theincreaseacceleratedin2008andisnowrisingfaster
thanatanypointsince1960.Thissharpandpredictableretirementofalargecohortcoincided
withthestartoftheGreatRecession,makingitdifficulttoseparatetheeffectsoftheeconomy
andnormaldemographicpatternsonparticipation.Eventhoughthesebabyboomersaremore
likelytoworkatolderagesthantheirpredecessors,theemploymentratesatolderagesarestill
muchlowerthanforyoungerworkers.


30
40
50
60
70
80
90
1948 1958 1968 1978 1988 1998 2008
Figure 2:MaleandFemale LaborForce
ParticipationRates
Percent
2014:Q2
Men
Women
10
15
20
25
30
35
40
1960 1970 1980 1990 2000 2010
Figure3:ShareoftheAdultPopulationOver
Age55
Percent
2013
55+
65+
9

II. Literature on the Labor Force Participation Rate in the Great


Recession and Subsequent Recovery

There has been considerable research on the decline in the participation rate in recent years.
Table1liststhefindingsofseveralrecentstudiesoftheparticipationrate,showinghowmuchof
the decline each study attributes to longterm trends and cyclical factors. The time periods
covered by each of these analyses vary, so the results are not always directly comparable
although we provide CEAs estimates to match the time periods of each of the studies. While
somestudiesestimateeitherthetrendorcyclecomponentandtheninfertheothercomponent
indirectly,thistableonlyshowsdirectestimatesofthesecomponents.

Despite differences in data and methods, several broad conclusions are common across the
literature. Taken together, the aging of the population and the business cycle account for the
vastmajorityofthedeclineintheparticipationratesince2007inallofthestudies.Whilethere
aresomedifferencesintheshareattributedtotheagingofthepopulation,themedianestimate
isthathalfofthedeclineisduetoaging.

Trend Cycl e Trend Cycl e


CEA(2014) 2007:Q42014:Q2 52% 16%
Begi nni ngi n2007
Erceg&Levi n(2013) 20072012 25% 60% 55% 35%
Shi erhol z(2012) 20072011 30% 50% 52%
VanZandweghe(2012) 20072011 42% 58% 50% 52%
CBO(2014) 2007:Q42013:Q4 50% 33% 49% 20%
Fal l i ck&Pi ngl e(2013) 2007:Q42013:Q2 75% 16% 55% 29%
Aaronsonetal .(2006) 20072013 80% 49% 20%
Kudl yak(2013) 20072012 80% 20% 55% 35%
OtherTi mePeri ods
Aaronson,Davi s&Hu(2012) 20002011 40% 46% 38%
Fuji ta(2014) 2000:Q12013:Q4 65% 30% 42% 16%
CEA(2014) 2011:Q12014:Q2 69% 34%
Table1:ComparisonofParticipationRateEstimates
Ti mePeri od
SharesoftheTotal Decl i ne
Note:Thi stabl eonl yi ncl udesdi rectesti matesofthetrendandcycl i cal components.CEAequi val entdecomposi ti ons
forti meperi odsl i stedonanannual basi sarecomputedfromQ4toQ4ofthecorrespondi ngyears.
CEAEsti matedSharesofTotal
Decl i neOverSameTi mePeri od

CEA estimates that the longterm trend of an aging population explains just over half of the
declineinparticipationsincetheendof2007,puttingCEAsestimateinthemiddleoftherange
ofestimates.Acrossmostofthesestudies,theeffectsofanagingpopulationtypicallyaccount
for a third to half of the decline, and the other longerterm trends account for a varying
proportion ofthedecline.TheestimatesofthetrendcomponentfromAaronsonetal.(2006),
Kudlyak(2013),andFallickandPingle(2013)alsoincludenotableeffectsfromotherlongterm
trends,andsoarecorrespondinglyabout25to30percentagepointslargerthanestimatesofthe
agingeffectsalone.
10

In analyzing the effects of the business cycle on participation, the literature is split between a
groupofanalysesthatattributeabout25percentofthedeclinetotheGreatRecessionandits
aftermath,andafewstudiesthatestimatethecyclicalcomponenttobeatleastamajorityofthe
decline. Each paper takes a different approach to estimating the cyclical portion. Both CEAs
estimate and Van Zandweghe (2012) are based on time series evidence but arrive at different
conclusions.

While some studies recognize an unexplained or residual component of the decline in


participation,othersholdthatthefulldeclineinparticipationcanbeexplainedbyacombination
oflongertermtrendsandcyclicaleffects.CEAestimatesthataboutonequarterofthedeclineis
not explained by either trend or historical estimates of cyclical movements, but there is some
evidence that this residual may be related to the current recession through its unusually high
longterm unemployment rate. The Congressional Budget Office (CBO) estimates a slightly
smallerresidualcomponentandsuggeststhatitmaybeduetoelevatedSocialSecurityDisability
Insurance(SSDI)rolls.Otherpossibleexplanationsfortheresidualcomponentincludeincreased
enrollmentinschool(Fujita2014).

11

III. Explaining the Decline in the Participation Rate Since 2007

CEAsanalysisofthedeclineinthelaborforceparticipationratesincetheendof2007isbased
onavarietyofmethodologies,includingdirectmeasurementoftheagingtrend,paneldata,and
timeseriesanalysisallofwhichproducegenerallysimilarresultsindisentanglingthreeeffects:
thepureagingeffect,thestandardcyclicaleffect,andaresidualwhichpicksupacombinationof
preexistingtrendsunrelatedtoagingandfactorsuniquetotheGreatRecession.

The Aging of the Workforce


Thelargestsinglefactorinthedeclineoftheparticipationratesincetheendof2007istheaging
of the workforcesomething that was predicted well before the Great Recession. Every year
since 2000, the Economic Report of the President has mentioned the post2008 decline in the
laborforceparticipationrateasafactorslowinggrowthofpotentialrealGDP.In2004,thefirst
year in which their tenyear forecast included the 20072013 period, CBO predicted that
participationwouldfallby1percentagepointbetween2007and2013duetoaging.Also,in2006
researchersfromtheFederalReservepredicted,basedsolelyonagingandothertrends,thatthe
participationratewouldfallto63.3percentin2013theexactvaluetowhichitfell(Aaronson
etal.2006).
Labor Force Participation over the Life Cycle
Laborforceparticipationvariesconsiderablyoverthecourseofapersonslife,risingrapidlyat
earlyagesbeforeflatteningoutthroughtheworkingyearsandfallingasworkersretire.Inorder
tounderstandhowanagingpopulationwillaffectoveralllaborforceparticipation,itishelpfulto
lookathowtheparticipationratesofdifferentagegroupshavechangedovertime.Figures4and
5showtheparticipationratesformenandwomenoverthecourseoftheirlivesforvariousbirth
cohorts.Forbothmenandwomen,participationrisesrapidlythroughoutthe20sasindividuals
leavefulltimeschoolingandentertheworkforce,thenflattensoutatahighlevelduringprime
workingageyears,beforefallingtolowlevelsofparticipationagainasindividualsentertheir60s.
Inthedecadeleadinguptoage 65,participationforboth menandwomenfallsbyatleast40
percentage points, and continues to fall by nearly 20 percentage points over the subsequent
decadeofthelifecycle.

12

Thepatternofparticipationamongprimeagewomenhasshifteddramaticallyoverthepasthalf
century. Women born in 1933 decreased their participation over the course of their 20sthe
agesofmaternityandchildrearingbeforeslowlyincreasingtheirparticipationintheir30sand
40s.Thecohortofwomenborntenyearslater,in1943,hadmuchthesamepatternofdecreasing
andincreasingparticipation,butoverthecourseoftheirlifetimeparticipatedatarateaboutfive
totenpercentagepointshigher.Insubsequentcohorts,womenparticipatedatmuchhigherrates
intheir20sand30s,whileparticipationroseonlyslightlyinolderyears.Mostofthisincreaseis
likely due to women in more recent cohorts returning to the labor force more quickly after
bearingchildrenthanpreviousgenerations.

0
10
20
30
40
50
60
70
80
90
100
21 26 31 36 41 46 51 56 61 66 71 76
1933
1943
1953
1963
Figure4:MaleLaborForceParticipationbyAge &
BirthYear
Percent
BirthYear
0
10
20
30
40
50
60
70
80
90
100
21 26 31 36 41 46 51 56 61 66 71 76
1933
1943
1953
1963
Figure5:FemaleLaborForceParticipation
byAge&BirthYear
Percent
BirthYear
13

Male participation has shown less variation from cohort to cohort, but there are two notable
changes.Thefirstisthatamongprimeworkingagemen,eachsuccessivecohorthasparticipated
slightlylessonaverage,demonstratingthatthelongrundeclineinmaleparticipationhasbeen
partially driven by the behavior of primeage men. The second is that more recent cohorts of
oldermenhaveincreasedtheirparticipation.Menbornin1943participatedatage65atarate
aboutfivetotenpercentagepointshigherthanmenbornin1933.Itisalsoworthnotingthat
menbornin1943hadalifeexpectancyroughly5yearslongerthanthosebornin1933.Despite
this increase in participation relative to previous cohorts, there is still a very large drop off in
participationatolderages,andsoanagingpopulationwill,allelseequal,reducetheeconomy
wide labor force participation rate. The substantial aging of the population that has begun to
occurandwillcontinuetounfoldforyearstocomeisdiscussedinthenextsubsection.
Aging and the Retirement of the Baby Boomers
In the immediate aftermath of World War II, the U.S. birth rate spiked, as couples who were
reunited following the war were able to settle down and have children, while others took
advantageoftherelativesenseofpeaceandprosperitythatemergedinthe1950sand1960sto
expand their families (Figure 6). The birth rate remained elevated into the mid1960s before
droppingsharply,andithasnotsincereturnedtoanywhereclosetowhatitwasduringthebaby
boom.Consequently,thebabyboomers,whoarecurrentlybetweentheagesof49and67,are
morenumerousthananypreviousorsubsequentcohortofAmericans.

AsFigure7shows,thedistributionofthepopulationbyagebecomesmoreheavilytiltedtoward
olderagesasthebabyboomersage.Thefirstbabyboomersreachedretirementagein2008,and
overthenextdecadeandahalf,increasinglylargecohortsofbabyboomerswillfollowsuit.In
2009,16percentoftheadultpopulationwasatorabovetheretirementage;theSocialSecurity
Administrationprojectsthisfractionwillgrowto25percentby2029,increasingbyhalfovertwo
decades.

12
16
20
24
28
32
1910 1930 1950 1970 1990 2010
Figure6:U.S.BirthRate
2013
Birthsperthousandpeople
14

Theretirementofthebabyboomgenerationisparticularlyconsequentialfortheparticipation
rate,becausethelaborforceparticipationratefallssharplyasworkersgrowolder.Bothmenand
womendecreasetheirparticipationbyatleast40percentagepointsinthedecadeleadingupto
age65,andparticipateatevenlowerratesafterwards.Becauseolderindividualsparticipatein
the labor force at lower rates than other age groups, the shift in the age structure from aging
babyboomersisresponsibleforasubstantialportionofthedeclineintheparticipationrate.

Although the participation rate of those over age 60 has risen in recent years and the
participationofolderbabyboomersiscontinuingtoriserelativetopreviouscohorts,thiseffect
isrelativelysmallcomparedtothetypicaldeclineaspeopleageandsodoeslittletomitigatethe
substantialnegativeeffectthatagingofthepopulationhasontheoverallparticipationrate.The
participationrateforthemalecohortbornin1943fellmorethan40percentagepointsasthey
agedfrom55to65,eventhoughtheirparticipationwasabout10percentagepointshigherthan
thoseinthepreviouscohortatage65.Thepatternforwomenislargelysimilar,astheincrease
inparticipationatolderagesfromcohorttocohortismuchsmallerthanthedropinparticipation
overthelifecycle.
Modeling the Effect of Aging on the Participation Rate
Twolongertermtrendscontributetomovementsintheparticipationrate:first,theeffectofan
aging population, and second, ongoing trends in participation within age groups that are
unrelatedtoaging.Themagnitudeofthetrendcanbeestimatedinseveralways,eachofwhich
captureslongtermmovementsinparticipationdifferently,withsomecapturingonlytheeffects
of aging and others including both aging and nonagingrelated trends. We employ three
different methods of modeling the participation rate trend and find that they provide similar
estimates,withapproximatelyhalfofthedeclineinparticipationsince2007attributabletoaging
andanegligiblefractionofthedeclineattributabletotrendswithinagegroups.

0
1
2
3
4
5
20 25 30 35 40 45 50 55 60 65 70
2019PopulationProjection
2013Population
Estimates
BabyBoomers
Figure7:U.S. PopulationDistributionByAge
Millions
15

ProjectingForwardthe2007ParticipationRateProfileWithChangingAgeStructure
A straightforward way to estimate the effect of the aging of the population on labor force
participationistoholdconstanttheparticipationratesbyageinagivenyear,andobservehow
theparticipationrateevolvesifonlytheagedistributionofthepopulationisallowedtovary.For
example, this allows computing the effect of an increasing share of 65 year olds, under the
assumption that 65 year olds retire in the future at the same rate that they did in 2007. This
approachisimplementedusingtheagebasedparticipationprofileasof2007.Theadvantageof
fixingparticipationratesin2007isthatitpredatedtherecession,andthusavoidsconfounding
theanalysisbyintroducingthefurthereffectofasharpcyclicalmovement.However,theresults
areessentiallythesameifweinsteadholdthecurrent(2014)participationratesbyagegroup
constant,asdiscussedlater.

To estimate the effect of aging, we constructed data on participation rates for singleyear age
categories by gender, combining unpublished participation rates for those over 55 with
participation for singleyear age categories below age 55 obtained by interpolating from the
published rates for tenyear age categories. For individuals above age 55, data on annual
participationbysingleyearagecategorywereprovidedtousbytheBureauofLaborStatistics
(BLS),usingunperturbeddatafromtheCurrentPopulationSurvey(CPS).Thesedataarelessnoisy
thanestimatesfromtheCPSmicrodatabutarenotverydifferentonaverage(AppendixFigures
1and2).

Figure 8 shows the counterfactual participation rate between 2007 and 2014, under the
assumption that only the age distribution of the population changed over this time. Over this
period,thecounterfactualparticipationratefellby1.6percentagepoints,accountingforslightly
more than half of the actual observed decline.
1
Thus, the aging of the population alone can
explainabouthalfthedeclineintheparticipationratesincelate2007.

1
Fixingtheparticipationratein2013andprojectingitbackwardsusingtheactualdemographicdatayieldsavery
similarestimatethat1.4percentagepointofthedeclineintheparticipationratewasduetoaging.Thisestimateis
slightlylowerthantheestimateusing2007databecausetheGreatRecessionnarrowedthegapinparticipation
betweenprimeageandolderworkers.
16

Theagingportionofthetrendhasbeenmorerapidlydecliningsince2011.Betweentheendof
2007andtheendof2010,thecounterfactualparticipationratefellonly0.6percentagepoint,
but it subsequently fell 1.0 percentage points from the beginning of 2011 through the first
quarter of 2014explaining the majority of the reduction in participation over that period. In
2011, the first baby boomers reached age 65, which is the point of steepest decline in
participationoverthelifecycle.Babyboomerswillcontinuetoturn65until2029andtheaging
portionofthetrendintheparticipationratewillcorrespondinglycontinuetodecline,possiblyat
anevenmorerapidpaceinthecomingyears.

Casselman(2014)conductsasimilarcalculationusingthepubliclyavailabledataonparticipation
published by BLS and concludes that aging is responsible for about a third of the decline.
However,theseasonallyadjustedstatisticspublishedbyBLScategorizealladultsover55intothe
same group, which obscures significant differences in participation at older ages. Participation
fallsbyabout40percentagepointsbetweenage55and65,socompositionalshiftswithinthe
55+ age group can have a large effect on the overall participation rate; using the coarse age
bracketdatawillnotcapturethiseffect.Incontrast,theanalysishereenablesanestimateofa
moredetailed,yearbyyearagingeffect.

TimeSeriesTrend
Econometricanalysisoftimeseriesdatacanalsobeusedtoestimatethelongruntrendinthe
participation rate. Unlike the previous approach that was designed to isolate the effect of
populationaging,thisapproachidentifiesasingleunifiedtrendintheparticipationtimeseries
andsodoesnotseparateagingandnonagingforcesdrivingthetrend.Althoughthetheoryof
timeseriestrendestimationiswelldeveloped,thisapplicationischallengingbecausethedeep
recessionandprolongedrecoveryservetoconflatethecyclicalandtrendbehaviorbeginningin
late 2007. For this reason, the time series trend used here is estimated using a standard time
series smoother (biweighted kernel), modified to be cyclically adjusted using a partially linear
semiparametricmodel;moredetailisgiveninAppendixA.Theresultingestimateshowsthatthe
62
63
64
65
66
67
2007 2008 2009 2010 2011 2012 2013 2014
Figure8:EffectofAgingonLaborForce Participation
Percent
AgingEffects
Actual
17

trendcomponentoftheparticipationratepeakedinthelate1990s,andhasbeenfallingsince,
decliningatamorerapidrateinrecentyears(AppendixFigure3).Accordingtothetimeseries
estimateofthetrend,thelongruntrendintheparticipationtimeseriesfellby1.6percentage
pointsbetweentheendof2007andthesecondquarterof2014.

The1.6percentagepointdeclineinthetimeseriestrendhappenstoequaltheestimateddecline
in the previous section that focused on the agingbased trend. Since the time series model
includestrendswithinagegroupswhiletheagingbasedtrenddoesnot,byitselfthisestimate
wouldsuggestthatnonagingtrendswithinagegroupshavehadlittleneteffectonparticipation.
However,inthepresenceofabreakintheserieswhichweknowoccurredaround2008with
the retirement boomtime series methods will tend to understate the trend. As a result, the
residualbetweenthetimeseriesandagingmethodswilltendtounderstatetheconsequences
ofpre2008trendsunrelatedtoaging.

StructuralMicroDataModelsofParticipation
Anotherapproachtoestimatingthetrendinparticipationinvolvesestimatingastructuralmodel
ofparticipationdecisionsusingindividualleveldata.Thisapproachwasoriginallydevelopedby
Aaronsonetal.(2006),whoextendthemodelwithageandcohorteffectstoincludecovariates
representinghumancapital,theabilitytofinancenonparticipationspells,andfamilystructure.
By including a measure of life expectancy, for example, the authors control for longer lives
requiringgreaterretirementsavings,whichinturnwouldincreaseparticipation.Incontrastto
the previous two methods, this approach attempts to ascertain the separate effect of various
factorsdrivingdifferencesinparticipationacrossindividualsandacrosscohorts.Aaronsonetal.s
projections,whichassumedfullemployment,wereconsistentwithadeclineintheparticipation
ratefrom2007:Q4through2014:Q2of2.5percentagepoints.Ifthisiscorrect,thenmostofthe
decline in participation since the end of 2007 would have been foreseeableand in fact was
almostpreciselyforeseenbasedonpreexistingtrendsinagingandotherdemographicfactors.

Fallick and Pingle (2007) propose a more detailed structural model of participation. They use
severalcovariatestoestimatethetrendinparticipation,includingeducationalattainment,life
expectancy,marriageandfertilityvariablesforwomen,minimumwagesandschoolenrollment
measuresforteenagers,andSocialSecurityrulechanges.Inarecentupdateoftheirmodelusing
datacollectedintheyearsfollowingpublication,theyestimatethattheparticipationratetrend
decreased2.0percentagepointsbetweentheendof2007andthesecondquarterof2014.Fallick
and Pingles model includes some portion of what we estimate as cyclical rather than trend
components, however,because several of thecovariates they includeexhibit cyclical variation
(in particular the return to education and the level of private wealth). To the extent that their
estimatedtrendcomponentisbasedonvariablesthatfluctuatecyclically,theirestimatearguably
overstatestheportionofthedeclineattributabletolongtermtrends.

18

Standard Business Cycle Effects


Thedeclineinthelaborforceparticipationratesince2007ispartiallyduetothestandardeffects
associatedwithaneconomyoperatingbelowitsfullpotential.However,becausetheshockof
theGreatRecessionwassolarge,thestandardcyclicaleffectsonparticipationwerealsoquite
large relative to previous cycles. When the economy is booming, the labor force participation
rate rises and often exceeds the predicted longterm trend, and during downturns the
participation rate falls below trend. Economic contractions historically result in both greater
unemploymentandlowerlaborforceparticipation,asnonparticipantsbecomelesslikelytoenter
thelaborforceandtheunemployed(whoalwaysexhibitahighertendencytoexitthelaborforce)
become more numerous relative to the employed (Elsby, Hobijn, and Sahin 2013). While
movementsintheparticipationrateoverdecadesaredrivenlargelybythelongtermtrends,in
the shortand mediumterm the variation in the participation rate is driven by these cyclical
factors.

Figure9showsthecyclicalityoftheparticipationratebycomparingthedetrendedparticipation
rate and the detrended unemployment gap, defined as the difference between the actual
unemployment rate and the natural unemployment rate (to simplify comparison, the
unemployment gap is plotted inverted)
2
. For example, in the expansion of the 1990s, the
detrendedparticipationraterose,andduringtheGreatRecession,thedetrendedparticipation
ratedeclined.Visualinspectionsuggeststhatmovementsintheparticipationratelag,byperhaps
a year or so, movements in the unemployment rate, a phenomenon discussed in more detail
below. The cyclical pattern of the participation rate is also evident when looking at statelevel
data,wherethestateswiththelargestincreasesinunemploymenttendedtohavelargerdeclines
inlaborforceparticipation(Figure10).Roughlyspeaking,foreverypercentagepointincreasein

2
Thetrendcomponentoftheparticipationratewasconstructedusingasemiparametricproceduredescribedin
AppendixA.Quarterlyaveragesofmonthlyparticipationwereusedtoreducethenoiseintheseries.
6
4
2
0
2
4
6
1960 1970 1980 1990 2000 2010
1.5
1.0
0.5
0.0
0.5
1.0
1.5
Figure9:DetrendedParticipationRateand
(Inverted)UnemploymentGap
Participation Rate
(leftaxis)
UnemploymentGap
(rightaxis)
PercentagePoints PercentagePoints
19

theunemploymentrate,theparticipationratefallsbyabout0.2percentagepoint,apatternthat
hasroughlyheldoverthelastfourrecessions(Figure11).


Whileitispossibletoestimateacyclicaleffectonparticipationusingmicrodata,thestudiesin
theliteraturethathavetriedthisapproachhaveproducedcyclicalestimatesthatdonotcomport
withthedynamicsevidentinthetotalparticipationtimeseries(Bengali,Daly,andValletta2013;
ErcegandLevin2013).Thecyclicalityofparticipationvariessignificantlyacrossthepopulation,
withparticipationamongyoungerworkersbeinghighlycyclicalandparticipationamongprime
ageworkershardlyrespondingtothebusinesscycle.Alternatively,onecanestimatethecyclical
componentindirectlybysubtractingthetrendcomponentfromtheobservedparticipationrate
andtreatingthisentireresidualasthecyclicalcomponent(VanZandweghe2012).However,this
6
5
4
3
2
1
0
0 2 4 6 8 10
PercentagePointChangeinUnemployment(20072010)
NV
CA
FL
IL
NY
TX
Figure10:Labor ForceParticipationand
UnemploymentbyState
PercentagePointChange inLFPR(20072012)
TrendLine
4.7
2.1
1.6
4.5
0.6
0.6
0.4
1.0
1.0
0.8
0.6
0.4
0.2
0.0 0
1
2
3
4
5
19791983 19901993 20012003 20072010
RiseinUnemployment
CyclicalEffectonParticipation
Figure11:CyclicalDeclines InLaborForceParticipation
PercentagePoints PercentagePoints
20

approachdoesnotdistinguishbetweenthetypicalcyclicalresponsefromasevererecessionand
the decline in participation coming from challenges unique to the Great Recession and its
aftermath.

Inthisreport,weestimatethecyclicalcomponentusingeconometricanalysisoftimeseriesdata,
wherethecycleisdefinedbytheunemploymentgap.Weconsidertwoalternativedistributed
lagspecifications,oneinwhichthedetrendedoverallparticipationrateisregressedonthelevel
of the detrended unemployment gap and its lags, and a second in which the quarterly first
differenceofthedetrendedoverallparticipationrateisregressedonthefourquarterdifference
ofthedetrendedunemploymentgapanditslags.Theregressionresultsforthetwomethodsare
summarizedinTables2and3,respectively.

VARIABLES (1) (2) (3) (4)


Unemp.Gap 0.0142 0.120** 0.165 0.140
(0.0615) (0.0530) (0.121) (0.126)
Unemp.Gap(t1) 0.216** 0.191*
(0.0957) (0.102)
Unemp.Gap(t2) 0.0184 0.0298
(0.0607) (0.0649)
Unemp.Gap(t3) 3.71e05 0.110
(0.0889) (0.0798)
Unemp.Gap(t4) 0.140*** 0.138 0.0451
(0.0428) (0.0853) (0.0686)
Unemp.Gap(t5) 0.0773
(0.0729)
Unemp.Gap(t6) 0.0652
(0.0622)
Unemp.Gap(t7) 0.0858
(0.0884)
Unemp.Gap(t8) 0.0462 0.00831
(0.0378) (0.0854)
Constant 0.0840 0.0619 0.0761 0.0821
(0.0870) (0.0902) (0.0863) (0.0878)
Observati ons 210 218 214 210
FtestofIncl udedCoeffi ci ents 8.530 5.143 6.075 4.006
Prob>F 2.29e05 0.0243 2.84e05 9.99e05
FtestofRestri ctedCoeffi ci ents 1.919 3.551 2.248
Prob>F 0.0793 0.000719 0.0652
Predi ctedCycl i cal Decl i ne(2007:Q4
2014:Q2)
0.477 0.113 0.465 0.486
Note:Regressi onsareesti matedusi ngdatafrom1960:Q1to2014:Q2.NeweyWest
standarderrorsusi ngamaxi muml agof12arereportedi nparentheses.
Parti ci pati onrateandunempl oymentgaparedetrendedusi ngtheprocedure
descri bedi nAppendi xA.Ftestsofrestri ctedcoeffi ci entsarethejoi ntsi gni fi cance
testoftheomi ttedcoeffi ci entsal l equal i ngzeroi naregressi onwi thal l ei ghtl ags
(i .e.i nregressi on(3),i tteststhehypothesi sthatthecoeffi ci entsonthe
unempl oymentgapi nt5throught8areal l equal tozero).
*** p<0.01, ** p<0.05, * p<0.1
DetrendedParti ci pati onRate,16+
Table2:LevelsonLevelsRegressionsusingthe
UnemploymentGap

21

VARIABLES (5) (6) (7) (8)


YearoverYearDi fferenceofUnemp.Gap 0.0340*** 0.0355*** 0.0105 0.0259
(0.00891) (0.00783) (0.0239) (0.0292)
YearoverYearDi fferenceofUnemp.Gap(t1) 0.0689 0.107*
(0.0500) (0.0551)
YearoverYearDi fferenceofUnemp.Gap(t2) 0.00142 0.0459
(0.0578) (0.0573)
YearoverYearDi fferenceofUnemp.Gap(t3) 0.0545 0.00622
(0.0468) (0.0536)
YearoverYearDi fferenceofUnemp.Gap(t4) 0.0187** 0.0473** 0.0630
(0.00740) (0.0204) (0.0514)
YearoverYearDi fferenceofUnemp.Gap(t5) 0.131**
(0.0637)
YearoverYearDi fferenceofUnemp.Gap(t6) 0.0899*
(0.0531)
YearoverYearDi fferenceofUnemp.Gap(t7) 0.0491
(0.0508)
YearoverYearDi fferenceofUnemp.Gap(t8) 0.00563 0.0179
(0.00751) (0.0393)
Constant 0.00141 0.00289 0.00371 0.00162
(0.00731) (0.00836) (0.00807) (0.00754)
Observati ons 210 218 214 210
FtestofIncl udedCoeffi ci ents 11.74 20.52 9.623 6.686
Prob>F 3.90e07 9.74e06 2.81e08 2.34e08
FtestofRestri ctedCoeffi ci ents 1.872 2.590 2.422
Prob>F 0.0873 0.0103 0.0496
Predi ctedCycl i cal Decl i ne(2007:Q42014:Q2) 0.507 0.303 0.544 0.474
Note:Regressi onsareesti matedusi ngdatafrom1960:Q1to2014:Q2.NeweyWeststandarderrors
usi ngamaxi muml agof12arereportedi nparentheses.Parti ci pati onrateandunempl oymentgap
aredetrendedusi ngtheproceduredescri bedi nAppendi xA.Ftestsofrestri ctedcoeffi ci entsarethe
joi ntsi gni fi cancetestoftheomi ttedcoeffi ci entsal l equal i ngzeroi naregressi onwi thal l ei ghtl ags
(i .e.i nregressi on(3),i tteststhehypothesi sthatthecoeffi ci entsontheunempl oymentgapi nt5
throught8areal l equal tozero).
***p<0.01,**p<0.05,*p<0.1
QuartertoQuarterDi fferenceofDetrended
Parti ci pati onRate,16+
Table3:DifferencesonDifferencesRegressionsusingthe
UnemploymentGap

SeveralresultsareevidentinTables2and3.First,inthespecificationswiththefullsetofeight
lags,manyofthecoefficientsareindividuallystatisticallyinsignificant.However,restrictingthe
lagstructuretoallow onlythecontemporaneousunemploymentratetoenter(regressions(2)
and (6)) is rejected at the 5 percent significance level. But once lags are introduced, some
restrictionsareconsistentwiththedata:oneofthespecificationswithshorterlags(3)andboth
specificationswiththecontemporaneous,fourth,andeighthlag((1)and(5))arenotrejectedat
the 5 percent level. Moreover, the predicted values from regressions (1), (4), (5), and (8) are
largely the same (Appendix Figure 4). Thus, the predicted results are largely robust across the
nonrejectedspecifications.

22

An important distinction between the firstdifference and levels specification is that the levels
specificationimposestherestrictionthatthelongtermforecastofthelaborforceparticipation
raterevertstoitslongtermtrend,whereasthedifferencesspecificationpermitsapersistentgap
betweenthelongtermtrendlineandtheparticipationrate(sothatineffectthetrendshiftsup
ordowntomatchpermanentshiftsintheleveloftheparticipationrate).Inthecurrentsituation,
thisdistinctioncorrespondstowhetherthegapbetweenthelaborforceparticipationrateand
its cyclical component (the residual) will entirely disappear, or whether that gap might persist
because of the labor market disruptions of the Great Recession. We view this as an empirical
question, and to avoid imposing the additional meanreverting structure on the cyclical
component,weadoptthefirstdifferencespecification.Further,toreduceestimationnoisewe
adopt the restrictedlag specification; this reasoning leads us to choose (5) as our preferred
specification. Note, however, that the decomposition results are not particularly sensitive to
thesechoicesbecausevaryingthemyieldssimilarestimatesofthecyclicalcomponents.

The predicted path for the participation rate, combining our estimate of the time series trend
with our preferred specification (regression (5) in Table 3), is shown below in Figure 12. We
estimate that between 2007 and 2014:Q2, business cycle effects alone explain 0.5 percentage
point(aboutonesixth)ofthetotalfallintheparticipationrate.

With the improvement of the labor market, the usual cyclical drag on participation due to
elevatedunemploymentisdeclining.Weestimatethat,asof2011,theparticipationratewas0.9
percentagepointlowerbecauseofcyclicalfactors.By2014:Q2,however,thecyclicalportionhad
shrunk to 0.5 percentage point because the unemployment rate has fallen. The drop in the
cyclicalcontributioncorrespondstonearly1millionworkersreturningtothelaborforcebecause
ofcyclicalreasonssince2011.
3

3
Duetoothersimultaneousflowsofworkersintoandoutofthelaborforce,thisreturnofworkerstothelabor
forcecanbedifficulttoidentifyintheaggregateddata.
62
63
64
65
66
67
68
1980 1990 2000 2010
Figure12:EstimatedTrendPlusCyclicalComponentof
theLaborForceParticipationRate,1980 2014
2014:Q2
Percent
Actual
Predicted CyclicalPath
(IncludingTrend)
23

Other Effects on Participation: NonAging Trends and Factors Unique to the


Great Recession
Mostofthedeclineinparticipationsincetheendof2007isaccountedforbytheagingofthe
population and the standard cyclical effects associated with the elevated unemployment rate.
The remaining reduction in the participation ratea 1.0 percentage point reduction in the
participation rate since the end of 2007is not explained by either of these two previously
discussedfactors.

Thisresidualisarelativelyrecentphenomenoninthisrecovery.Upuntilthebeginningof2012
theparticipationratewasgenerallyslightlyhigherthanwouldhavebeenpredictedbasedonthe
agingtrendandthestandardbusinesscycleeffects.Butinthelasttwoyears,theparticipation
ratehascontinuedtofallataboutthesamerateeventhoughtheunemploymentratehasbeen
decliningrapidly.Thismeansthatthestandardcyclicaleffecthasbeengivingwaytoagrowing
unexplainedresidual(Figure13).

Theliteraturehasputforwardseveralpossibleexplanationsforthisresidualcomponentwhich
fall mainly into two categories: longterm trends within age groups and factors unique to the
GreatRecession.Wecannotpreciselybreakdownthesetwofactors.Atleastsomeoftheresidual
appearstobethesepreexistingtrends,manyofwhicharediscussedinthenextsectionabout
differentdemographicgroups,althoughsomeofthesepreexistingtrendsarepartiallyoffsetting
(e.g., primeage male labor force participation was already falling while older workers were
seeingarisingparticipationrate).

Thefactthattheresidualonlyemergedlaterintherecoverysuggeststhatmuchofitmightbe
theresultoffactorsspecifictotheGreatRecessionanditsaftermath.Wefocushereontesting
threeuniquefactorsoftheGreatRecessionthatmaybelinkedtotheresidualcomponent,but
otherforcesmayalsoexplaintheresidualdecline.

24

Significantly Elevated LongTerm Unemployment


TheGreatRecessionhasledtoanunprecedentedincreaseinlongtermunemployment,withthe
number of individuals unemployed for 27 weeks or more rising from 1.3 million at the end of
2007 to more than 6 million at the end of 2010. Even though the rate of shortterm
unemployment (26 weeks or less) has since returned to its 200107 average, the longterm
unemploymentrateremainsmorethandoublewhatitaveragedduringthepreviousexpansion.

The emergence of a large residual in our participation rate model coincides with a sustained
elevationofthelongtermunemploymentrate,suggestingthatthesetwophenomenacouldbe
differentmanifestationsofthesameunderlyingchallenge,orthatoneiscausingtheother.To
testtheplausibilityofthishypothesis,welookatthehistoricalrelationshipbetweentheresidual
(unexplained component) and the extent of longterm unemployment. Specifically, we ran
several regressions using the detrended mean duration of unemployment as a regressor in
additiontothedetrendedunemploymentgapthatwasusedinthestandardcyclicalestimates
above.
4
AsTables4(levelsonlevels)and5(differencesondifferences)show,themeanduration
of unemployment was highly statistically significant in the differences specification but only
marginallysignificantinthelevelsspecification.Thesetablesalsoshowtheeffectsofdisability
insuranceandschoolenrollments,whichwillbediscussedbelow.

4
TheCurrentPopulationSurveywaschangedatthebeginningof2011topermitrespondentstoreportlonger
durationsofunemployment.Thischangemayhaveledtoabreakintheleveloftheseries,whichmightinfluence
ourregressionsinlevels.Thechangeisnotlikelytoaffectourregressionindifferences,whichiswherewefind
thatmeandurationissignificant.Usingthemediandurationofunemployment(whichwasnotaffectedbythe
change),wefindthatunemploymentdurationissignificantinthelevelsspecifications,butnotthedifferences
specifications,providingfurtherevidencethatlongtermunemploymentaccountsfortheresidualcomponent.
25

VARIABLES (1) (2) (3) (4) (5) (6)


Unempl oymentGap 0.0142 0.0712 0.0130 0.0154 0.00373 0.0153
(0.0723) (0.0805) (0.0574) (0.0597) (0.0619) (0.0640)
Unemp.Gap(t4) 0.140*** 0.169** 0.142*** 0.137*** 0.137*** 0.135***
(0.0525) (0.0654) (0.0417) (0.0455) (0.0431) (0.0404)
Unemp.Gap(t8) 0.0462 0.0249 0.0512 0.0499 0.0332 0.0453
(0.0609) (0.0569) (0.0425) (0.0453) (0.0402) (0.0468)
MeanDurati onofUnemp. 8.35e06 0.0400
(0.0224) (0.0279)
MeanDurati onofUnemp.(t4) 0.0356
(0.0296)
MeanDurati onofUnemp.(t8) 0.0119
(0.0242)
SSDICl ai ms 0.000933** 0.000615
(0.000468) (0.00115)
SSDICl ai ms(t4) 0.000272
(0.00188)
SSDICl ai ms(t8) 0.000121
(0.00124)
Enrol l mentRate 22.92 41.04
(19.18) (32.51)
Enrol l mentRate(t4) 5.962
(27.62)
Enrol l mentRate(t8) 26.26
(39.03)
Constant 0.0840 0.0842 0.0860 0.0859 0.0997 0.0892
(0.0892) (0.0900) (0.0881) (0.0881) (0.0799) (0.0792)
Observati ons 210 210 204 204 210 210
Ftest 1.38e07 2.624 3.974 0.157 1.429 0.379
Prob>F 1.000 0.0750 0.0476 0.855 0.233 0.685
Table4:LevelsonLevelsRegressionsusingDisabilityClaims,DurationofUnemployment,and
EnrollmentRate
DetrendedParti ci pati onRate,16+
***p<0.01,**p<0.05,*p<0.1
Note:Regressi onsareesti matedusi ngdatafrom1960:Q1to2014:Q2.NeweyWeststandarderrorsusi ngamaxi muml agof12arereported
i nparentheses.Parti ci pati onrateandunempl oymentgaparedetrendedusi ngtheproceduredescri bedi nAppendi xA.Ftestsarejoi nt
si gni fi cancetestsofthedi sabi l i tyi nsurance,meandurati on,andschool i ngvari abl es.

26

VARIABLES (7) (8) (9) (10) (11) (12)


YearoverYearDi fferenceofUnemp.Gap 0.0297*** 0.0330*** 0.0322*** 0.0321*** 0.0330*** 0.0328***
(0.00890) (0.00914) (0.00989) (0.00992) (0.00884) (0.00877)
YearoverYearDi fferenceofUnemp.Gap(t4) 0.0130 0.00429 0.0228** 0.0149 0.0181** 0.0190**
(0.0103) (0.0146) (0.00943) (0.0111) (0.00724) (0.00737)
YearoverYearDi fferenceofUnemp.Gap(t8) 0.00350 0.0151 0.0117 0.0170* 0.00481 0.00569
(0.00776) (0.0114) (0.00916) (0.00901) (0.00786) (0.00744)
YearoverYearDi fferenceofMeanDurati on 0.00387 0.00406
(0.00436) (0.00534)
YearoverYearDi fferenceofMeanDurati on(t4) 0.0142***
(0.00524)
YearoverYearDi fferenceofMeanDurati on(t8) 0.00222
(0.00527)
YearoverYearDi fferenceofSSDICl ai ms 0.000546 0.000610
(0.000614) (0.000874)
YearoverYearDi fferenceofSSDICl ai ms(t4) 0.00120
(0.00117)
YearoverYearDi fferenceofSSDICl ai ms(t8) 0.000719
(0.000756)
YearoverYearDi fferenceofEnrol l mentRate 2.510 2.910
(4.275) (4.702)
YearoverYearDi fferenceofEnrol l mentRate(t4) 2.876
(4.415)
YearoverYearDi fferenceofEnrol l mentRate(t8) 0.444
(3.391)
Constant 0.000523 0.00247 0.00154 0.00225 0.00124 0.00147
(0.00743) (0.00676) (0.00658) (0.00698) (0.00732) (0.00719)
Observati ons 210 210 145 137 210 210
Ftest 0.789 3.529 0.788 0.373 0.345 0.161
Prob>F 0.376 0.0159 0.376 0.773 0.558 0.923
Note:Regressi onsareesti matedusi ngdatafrom1960:Q1to2014:Q2.NeweyWeststandarderrorsusi ngamaxi muml agof12arereportedi nparentheses.
Parti ci pati onrateandunempl oymentgaparedetrendedusi ngtheproceduredescri bedi nAppendi xA.Ftestsarejoi ntsi gni fi cancetestsofthedi sabi l i ty
i nsurance,meandurati on,andschool i ngvari abl es.
***p<0.01,**p<0.05,*p<0.1
QuartertoQuarterDi fferenceofDetrendedParti ci pati onRate,16+
Table5:DifferencesonDifferencesRegressionsusingDisabilityClaims,DurationofUnemployment,and
EnrollmentRate

Usingtheseregressionresults,wecanpredictthepathoftheparticipationrategiventhelarge
increaseinthemeandurationofunemploymentsince2007.Estimatingthisnewsetofadditional
factorsexplainsmuchofthepreviouslyunexplaineddeclineinparticipationsince2009,asseen
in Figure 14. This suggests that the unusually high rate of longterm unemployment may be
exerting downward pressure on the participation rateor that these two developments are
manifestationsofacommonunderlyingcause.

27

Onetheoryputforwardtoexplaintheriseinlongtermunemploymentduringthemostrecent
recessionisthatthelongtermunemployedarelessproductiveworkersonaverageandwilltake
longertofindnewjobs(CowenandLemke2011).However,thelongtermunemployedappear
similar to the shortterm unemployed in many dimensions, including gender, educational
attainment, and previous industry and occupation (Mitchell 2013; Krueger, Cramer, and Cho
2014).Lowerproductivityamongthelongtermunemployedcouldberelatedtounobservable
characteristics,butitis notclearwhetherthedifferencesamongunobservabletraitsarelarge
enoughtoexplaintheriseinlongtermunemployment.

Amoreplausibleexplanationfortheriseinlongtermunemploymentisthatfirmsarescreening
outapplicantsbasedondurationofunemployment.Longtermunemployedworkersareabout
half as likely to get a callback for an interview as equallyqualified shortterm unemployed
workers (Kroft, Lange, and Notowidigdo 2013). Shortterm unemployed applicants with no
relevantworkexperienceforaparticularjobaremorelikelytobeinterviewedforthejobthan
longtermapplicantswithrelevantworkexperience(Ghayad2013).Workerswhoarenotlucky
enoughtofindajobwithinthefirstfewmonthsofunemploymentwillfacedisadvantagesinthe
hiringprocess,furtherlengtheningtheirunemploymentspellsonaverageandraisingtherateof
longtermunemployment.

Recent research finds that the period of elevated longterm unemployment that we are
experiencing is consistent with historical patterns after adjusting for the deeper decline and
slowerrecoveryfromthelatestrecession(Kroft,Lange,Notowidigdo,andKatz2014).Consistent
with the fact that the longterm unemployed are observably similar to the shortterm
unemployed,Kroftetal.findthatdemographics,occupations,industries,regionsorreasonsfor
unemployment do little to explain the variation in unemployment spell length. Instead, they
attributealargeshareoftheriseinlongtermunemploymenttonegativedurationdependence,
meaningthatasaspellofunemploymentpersists,itbecomesmorelikelytopersistevenlonger.
Theyfurtherfindthatchangestheflowsbetweenunemploymentandnonparticipationinthe
62.5
63.0
63.5
64.0
64.5
65.0
65.5
66.0
66.5
67.0
2009 2010 2011 2012 2013 2014
Figure14:LaborForce ParticipationDecomposition
IncludingMeanDurationofUnemployment
Percent
RemainingResidual
OtherEffects
MeanDuration
Actual
28

laborforceaccountforanotherlargeshare.Theyallowthatthelongtermunemployedmaybe
unobservablylessproductive,buttheirtwofavoredexplanationsleavelittleunexplained.
Disability Insurance
Since the end of 2007, the number of disabled workers receiving benefits through the Social
SecurityDisabilityInsurance(SSDI)programhasincreasedbynearly2million.Somehaveargued
that changes in Disability Insurance are causing an unusual decline in the participation rate.
However,wefindlittleevidencethattheincreaseinDisabilityInsurancebeneficiariesexplains
theresidualcomponentofthedeclineinparticipation.Weestimatedaseriesofregressionsusing
thedetrendedtimeseriesofDisabilityInsurancedisabledworkerbeneficiariesasanadditional
explanatory variable. Tests of the coefficients were mostly statistically insignificant and only
barelysignificantinonespecification(regression(3)inTable4).Lookingatthepredictedvalues
from theseregressions, the estimated path of participation accounting for Disability Insurance
didnotfitthehistoricalseriessubstantiallybetterthanourpreferredspecificationusingonlythe
unemploymentgap.

Disability Insurance generally increases when the unemployment rate increases. As such, the
increaseinDisabilityInsurancebeneficiariesiscapturedinourestimateofthecyclicaleffectand
does not appear to have had a disproportionate effect in recent years. In terms of this
decomposition, Disability Insurance does not appear to explain the residual component of the
declineintheparticipationrate.
5

Infact,ifanything,theincreaseindisabilityrollssince2009havebeensomewhatlowerthanone
would have predicted given the predicted cyclical and demographic effects, although it is still
withina95percentconfidenceinterval(Figure15).

5
Figure15showsthepredictedlevelofSSDIbeneficiariesusingthesameapproachfordoingtimeseries
decompositionasouranalysisoftheparticipationrate.WeregressedfirstdifferencesinthedetrendedlogofSSDI
beneficiariesataquarterlylevelonfourquarterdifferencesinthedetrendedunemploymentgap.Usingtheactual
numberofbeneficiariesattheendof2007asaninitialpoint,wethencumulatedthedifferencesandthe
estimatedconfidenceintervalusingNeweyWeststandarderrorstoproduceanestimatedrangeforthenumberof
disabledworkerbeneficiaries.
29

Schooling
Youngmenandwomenareincreasinglyenrollinginschoolfulltimeinsteadofenteringthelabor
force. The effects of the Great Recession have further encouraged young workers to enroll in
school and delay entering the labor force. However, school enrollment tends to be counter
cyclicalandthereforeanyeffectofatypicalincreaseinschoolingduringarecessionisalready
includedinourestimateofthecyclicalcomponent.Iftheincreaseinschoolenrollmentbetween
2007and2014islargerthanthetypicalincreaseduringarecession,though,itmayexplainsome
oftheresidualdeclineinparticipation.

Weestimateseveralregressionspecificationsinbothlevelsanddifferencesusingthepercentof
thepopulationenrolledinhighschoolorcollegeasanadditionalexplanatoryvariable.AsTables
4 and 5 show, theenrollment rate is not statistically significant in any of the specifications we
estimate.Analyzingthecyclicalpatternsintheenrollmentrateitselfrevealsthatthedynamics
of enrollment between 2007 and 2014 are largely in line with the expected path given the
countercyclicalnatureofenrollmentinthepast.

It is important to note that this analysis only captures the effect of changes in enrollment
decisions,andnotchangingparticipationconditionalonenrollment.Studentsenrollinginschool
participate at lower rates than their nonenrolled peers, so a higher enrollment rate leads to
lower participation on average. However, the participation of students enrolled in school has
been on a downward trend since 2000, which accounts for a portion of the decline in
participation among younger workers. This implies that the participation decisions of young
workersmayhavechangedinwaysthatarenotfullycapturedintheenrollmentrate.However,
whenlookingattheoverallparticipationrate,thischangeispartoftheotheroffsettingtrends
discussedingreaterdetailbelow.

6
7
8
9
10
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Figure15:PredictedSSDIBeneficiaries
Actual
Beneficiaries (Millions)
Predicted
Upper/Lower
95%Bounds
30

IV. Changes in Labor Force Participation for Different Demographic


Groups

Overall,themajoraggregatetrendaffectingtheparticipationrateintherecessionandrecovery
istheagingofthepopulation.Butthereareanumberofimportanttrendsanddevelopmentsfor
different subgroups of the population. For example, we have seen declining participation by
youngerAmericansastheystayinschoollonger,increasingparticipationbyolderAmericans,a
plateauingof,andsubsequentslightdeclinein,femalelaborforceparticipationinthelastfifteen
yearsfollowingearlierincreases,andthecontinuationofa65yeartrendofdecliningmalelabor
forceparticipation(whichisespeciallystarkforyoungminoritymen).

Table6showsthetrendsforawiderangeofgroupsandtheremainderofthissectiondiscusses
trendsforsomeofthespecificgroups.

19481990 19902007 20072014


All 62.8 0.2 0.0 0.5
Men 69.1 0.2 0.2 0.6
Women 56.9 0.6 0.1 0.3
1624 54.7 0.2 0.4 0.8
2554 80.8 0.5 0.0 0.3
55+ 40.0 0.3 0.5 0.2
Men2554 88.0 0.1 0.1 0.5
Women2554 73.9 1.0 0.1 0.2
White* 63.1 0.2 0.0 0.5
Black** 60.9 0.3 0.0 0.5
Hispanic*** 65.8 0.5 0.1 0.5
BlackMen** 63.3 0.1 0.2 0.5
BlackWomen** 58.9 0.6 0.1 0.4
*Seriesbeginsin1954
**Seriesbeginsin1972
***Seriesbeginsin1973
Table6:ParticipationRatebySelectedGroups
2014Q2
Change(peryearinp.p.)

Young Men and Women


Labor force participation among young workers (1624 year olds) is at its lowest level in more
than50years,havingfallensteadilysinceitspeakinthelate1980s,andisnow15percentage
pointslowerthanitwas30yearsago,asseeninFigure16.However,giventherelativelysmall
size of this group, this trend has had a much smaller impact on the overall labor force
participation rate. Since 2007, participation among young people has fallen by about 0.8
percentagepointperyear,whileschoolenrollmenthasincreased.Enrollmenthasrisensincethe
1980swithlargerincreasesinenrollmentamonglowerincomehouseholds,asshowninFigure
31

17.Asaresult,theshareofyoungpeoplewhoareeitherinthelaborforceorinschoolhaslargely
beenunchanged.

Aaronson, Park, and Sullivan (2006) examine the secular decline in teen (1619 year olds)
employmentdatingbacktothemid1980sandhypothesizethatteensarerespondingtotherise
in returns to college by enrolling in school. They find that most of the decline in employment
between1987and1997wasduetoariseinschoolenrollment,butsincethenmostofthedecline
is due to reduced participation among young people enrolled in school. Over the last two
decades, the share of 1624 year olds neither enrolled in school nor participating in the labor
force has not changed significantly, ticking down slightly from 11.9 percent in 1993 to 11.7
percentin2013.Thelackofanyincreaseintheshareneitherenrollednorparticipatingimplies
50
60
70
80
90
100
1948 1958 1968 1978 1988 1998 2008
Figure16:LaborForceParticipationRateof
1624YearOlds
2014:Q2
Percent
ParticipationRate
Share EitherEnrolledinSchoolor
ParticipatingintheLaborForce
(4Qtr.MovingAverage)
32

thatthedeclineinthe1624yearoldparticipationratesince1993hascomefromareductionin
participationamong1624yearoldsenrolledinschool.Indeed,theparticipationrateof1624
year olds enrolled in school has declined more than 10 percentage points over the past two
decades.

Two main factors have likely played a role in the increase in enrollment since 1980, which
accounted for much of the decrease in participation among the young between 1980 and the
mid1990s.First,increasedsubsidiesforcollegecombinedwithhigherreturnstostayinginschool
havechangedtheeconomiccalculusaroundschoolingdecisions.Second,areductioninmiddle
skilled jobs (which economists have referred to as job polarization) has led lessskilled adult
workerstopushteenageworkersoutofthelaborforce.

Smith (2011) examines the decline in the teen (1619 year olds) employmentpopulation ratio
whichhasbeentrendingdownsincethemid1980s,steepeningintherecentperiodsinceearly
2000s. Using crossstate analysis, the author finds a significant negative relationship between
changesinteenemploymentandthefractionofadultsinjobsthatoftenemployteenssuchas
lowwage service sector jobs. Job polarizationdeclining middleskilled jobs and growing low
wagejobsmayhaveresultedinincreasedcompetitionbetweenlessskilledadultworkersand
teenage workers. The author finds that such explanations can account for over half of the
cyclicallyadjusted decline in teen employment since the mid1980s. It is plausible that these
forces have grown over the recent recession as youth, particularly teens, faced greater
competition for jobs from more experienced adults who were laidoff (Fogg and Harrington
2011).

Older Workers
After many decades of decline, labor force participation rates among older workers have
stabilizedsincethe1980sandhavestartedtoincreaseinrecentyears(Figure18).Participation
amongolderwomenwaslargelyflatforseveraldecadesbeforeexperiencinganunprecedented
increaseoverthelasttwodecades,risingmorethan10percentagepoints.Thisriseinthelabor
force participation for older workers has been attributed to several factors. Maestas and
Zissimopoulos(2010)andBlauandGoodstein(2010)identifyrisingeducationlevels,laborforce
participationofwives,andchangesinSocialSecurityasthemostimportantfactorsleadingtothe
reversal in the trend. While there is less direct evidence available, it seems likely that
improvementsinhealthandincreasesinlifeexpectancyhavealsoplayedarole.

33

Theincreaseinparticipationbyolderworkersispartiallyduetodifferencesintheskillsthatthey
acquiredwhentheywereyounger.Between1876and1950,theaverageyearsofschoolingfor
each birthyear cohort increased steadily every year, according to Goldin and Katz (2007). The
increasededucationamongeachsuccessivecohortmeantthatretireesinmorerecentyearshave
hadmoreincomeandbetterquality,lessphysicallydemandingjobsoverthecourseoftheirlives
(MaestasandZissimopoulos2010).BlauandGoodstein(2010)estimatethatatleast15percent
oftheincreaseinlaborforceparticipationsince1988canbeattributedtotherisingeducation
levels in each successive cohort, but their robustness checks indicate that this may be an
underestimate and rising education might in fact explain the entire increase in participation
amongolderworkerssince1988.

Another portion of the recent rise in participation among married men has come from the
increased labor force participation of their wives. The rise in female labor force participation
sincethelate1980sexplainsaboutaquarteroftheriseamongoldermen(BlauandGoodstein
2010,Schirle2008).Thiseffectislikelyduetocomplementarityofleisuretimebetweenhusbands
andwives,whereretirementismoreappealingifonesspouseisalsoretired.Couplestendto
retireatsimilartimesandmensretirementdecisionsareveryresponsivetotheincentivesfaced
bytheirwives(Coile2004).

Additionally,therehavebeenanumberofpolicychangestotheSocialSecurityprogramsince
1980,andthesepoliciesmayhavehadaneffectontheretirementdecisionsofolderworkers.In
particular,researchinthisareahascenteredaroundtwotypesofpolicies:changingthelevelof
fullbenefitsandchangingthetimingofwhenfullbenefitsarepaidout.Changesinthegenerosity
ofbenefitsappeartohavehadlittleeffectonthelaborsupplyofolderworkers,whilechangesin
thetimingoffullbenefitsaccountformuchoftheincreaseinparticipationatolderages(Blau
and Goodstein 2010, Gustman and Steinmeier 2008). In particular, the increase in the Normal
Retirement Age and the creation of the Delayed Retirement Credit (that is, the increase in
benefitsfromworkingpastthenormalretirementage)togethercanexplainbetweenaquarter
25
30
35
40
45
1948 1958 1968 1978 1988 1998 2008
Figure18:LaborForceParticipationRateof
55+YearOlds
2014:Q2
Percent
34

andhalfoftheincreaseinlaborforceparticipationamongolderworkerssince1988(Blauand
Goodstein2010).

Thereappearstobe,asyet,littleinthewayofrigorousempiricalstudiesexaminingtwofactors
whichlikelycontributedtotherecentriseinparticipationamongolderworkers:improvements
inhealthandtechnology.Improvementsinhealthcanenableworkerstocontinuetoworklater
in life and the corresponding increase in longevity will raise the amount of saving needed for
retirement.Additionally,technologicalimprovementscanalsohelpaccommodateolderworkers
intheworkplace.

As older workers have delayed retirement and increased their participation in the labor force,
therehasbeenariseinpartialretirementasameansofshiftinggraduallyfromthelifestyleof
primeworkingyearstoretirement(Maestas2007).AccordingtoMaestas,mostoftheseretirees
anticipated working in retirement, and retirees who unexpectedly return to the labor force
typically do so becausethey find leisure less enjoyable than theyexpected. Additionally, older
workersareincreasinglylikelytobeselfemployed(MaestasandZissimopoulos2004).About25
percentofall65yearoldworkersareselfemployed,andoftheseworkersmorethan40percent
haveemployeesworkingforthem(MaestasandZissimopoulos2010).

Between 1990 and 2007, participation among workers over 55 grew by an average of 0.5
percentagepointperyear,butthisslowedto0.2percentagepointperyearsince2007(Table6).
Ouranalysisoftheover55participationratesuggeststhattheincreasesince2007wasnotdriven
by the Great Recession, and in fact we estimate that the cyclical downturn actually slightly
decreased the participation of older workers relative to what it would have otherwise been.
ResearchontheeffectsoftheGreatRecessionislargelyinlinewiththisresult,findingthatwealth
shocks were concentrated in wealthier households, and as a result did not substantially alter
retirementdecisionsinrecentcohorts(Gustman,Steinmeier,andTabatabai2011).

PrimeAge Women
Thelaborforceparticipationrateforprimeagewomenroserapidlyfromaround1970through
2000.Thisrapidrisewasdrivenbymanyfactorscomingtogether:reductionsindiscrimination
facilitatedbyTitleVIIoftheCivilRightsActandTitleIXoftheEducationalAmendmentstothe
Civil Rights Act, the availability of more effective birth control through the invention and
subsequentlegalizationofthebirthcontrolpill,andchangesineconomicforcesstemmingfrom
thegreatereducationalgainsofwomen,theadventoflaborsavinghouseholdtechnology,and
increasedinternationaltrade.Whilethesechangesdroveparticipationinthe1970sand1980s,
further increases in the 1990s stemmed from the incentives provided by the expansion of the
Earned Income Tax Credit (EITC), the replacement of Aid to Families with Dependent Children
(AFDC)withtheworkcenteredTemporaryAssistanceforNeedyFamilies(TANF),andreformsto
divorcelaws.Buttheparticipationratehassincelevelledout,asseeninFigure19.Asaresult,
the United States has gone from leading many other European countries in primeage female
laborforceparticipationtolaggingthem(Figure20).

35

Research has found that familyfriendly policies are partially responsible for the rise in
participationinotheradvancedcountries,andthelackofthesepoliciesexplainswhytheUnited
States has lost ground. Policies such as paid parental and sick leave, high quality available
daycare, and support for parttime work have been enacted in other advanced countries,
enabling mothers to more easily balance work and family obligations. Blau and Kahn (2013)
investigateddataonparticipationacrossdevelopedcountriesandfoundthatthelackoffamily
friendly policies in the United States can explain roughly a quarter of the relative decline in
womensparticipationbetween1990and2010.

Accesstopaidfamilyleaveisaparticularlyimportantfactorformothersconsideringwhetheror
nottoremaininthelaborforce,sinceitenablesmotherstocontinueworkingforanemployer
after giving birth. In 2004, California became the first State to implement a paid family leave
35
45
55
65
75
85
1948 1958 1968 1978 1988 1998 2008
Figure19:LaborForceParticipationRateof
2554YearOldWomen
Percent
2014:Q2
36

program.RossinSlater,Ruhm,andWaldfogel(2013)examinedthedecisionsofmothersbefore
andaftertheprogramsimplementation,findingthatitsignificantlyboostedthenumberofhours
that mothers worked two to three years after giving birth. They also find that the program
reduced the number of mothers without jobs, although this effect was not quite statistically
significant.Otherresearchonpaidleaveprogramshasshownthataccesstopaidleaveincreases
thelikelihoodthatarecentmotherwillreturntoheremployerintheUnitedStates,theUnited
Kingdom,andJapan(Waldfogel,Higuchi,andAbe1999),althoughastudyofpaidleaveinNorway
found no effect on labor force participation, perhaps reflecting stronger preexisting work
supports(Dahletal.2013).

Atthesametimeastheparticipationrateforwomenhasstagnated,theroleofwomeninthe
workforce has continued to grow in importance. More and more women are working in
traditionally maledominated occupations and becoming breadwinners for their families:
married women now contribute almost half of family earnings. This is in part because young
womenarenowmorelikelythanyoungmentohavegraduatedfromcollege,andtheshareof
femalestudentsenrollinginMD,JD,andMBAprogramshasgrowntonearly50percentforeach
ofthesedegrees.

PrimeAge Men
Therehasbeenasteadydeclineinthelaborforceparticipationofprimeagemensincethe1950s
(Figure21).In1950,92percentofprimeagemenwereemployed,whilethemostrecentdata
showthatthepercentagehasfallento83percent,inotherwordsadeclineofnearly10percent.
Thisdeclineinpartreflectsthefactthatprimeagemenhavehistoricallybeenmoreconcentrated
in jobs that can be replaced by technology or sent overseas. As described in Goldin and Katz
(1998)andAutor,Katz,andKearney(2008),skillbiasedtechnologicalchangehaslongexerted
aninfluenceonthewagestructurebydepressingwagesforoccupationswithmanualorroutine
labor.AsAutor,Dorn,andHanson(2013)note,theseeffectsareconcentratedamongindustries
exposedtoforeigncompetitionandinternationaltrade.

37

There is some disagreement in the literature on whether SSDI accounts for a part of the
participation decline, particularly for lowerskilled male workers. Wage stagnation for less
educated workers have made Social Security benefits increasingly valuable compared to low
wage workers earnings, and reforms passed in 1984 increased SSDI receipt, although this
increase may have reflected the fact that the level of SSDI receipt was temporarily depressed
before 1983 and subsequently returned to its normal level after reforms were enacted (Autor
andDuggan2003,PattisonandWaldron2013).Asdiscussedabove,morerecentgrowthinthe
programhasbeendrivenmostlybydemographicsandtherecession.

Regardless of the source, the prolonged decline in participation among primeage men is
worryingbecausethisgrouphashistoricallycomprisedthelargestshareoftheworkforce.Even
seeminglymodestdeclinesintheirparticipationcumulateoverlongperiodsoftimeandresultin
largechangesinthenumbersofAmericansavailableforwork,andthisinturnhaspredictable
negativeeffectsongrowth.

Black Men
Blackmalelaborforceparticipationhasfallensteadilysincethemid1970s,droppingmorethan
10 percentage points over that time. In part, this reflects lower education levels among black
men,only47percentofwhomtodayhaveatleastsomecollegeeducation(comparedwith57
percentforwhitemen).Forbothblackandwhitemen,laborforceparticipationhasfallenmore
for those with lower education levels. At the same time, however, the decline in participation
has been greater for black men across the education spectrum. Participation among black
workers with just a high school degree has declined nearly 10 percentage points over the last
decade, while the participation rate among black college graduates has declined by 3.5
percentagepoints;thiscompareswith4.6percentagepointsand2.7percentagepointsdeclines
forwhitemen,respectively.Thistroublingtrendofreducedparticipationhascontinuedinthe
85
87
89
91
93
95
97
99
1948 1958 1968 1978 1988 1998 2008
Figure21:LaborForceParticipationRateof
2554YearOldMen Percent
2014:Q2
38

latest business cycle, with the participation rate for black men down 0.5 percentage point per
yearsincetheendof2007(Figure22).

The decline in black male participation, which has persisted through both recessions and
recoveries,haslongbeenthesubjectofresearch(Moynihan1965,BoundandFreeman1992).
Thedeclinehasalsobeenparticularlypronouncedamongtheyoung:whiletheparticipationrate
ofolderblackmenisalsobelowaverage,theirsishigherthanthatofyoungblackmen,whohave
a lower labor force participation rate than young white and Hispanic men. This persistent gap
between black men and their peers in participation and employment opportunities more
generallyislikelyduetoarangeoffactorssuchasdisparitiesinfamilylifeandeducation,aswell
as employer attitudes and the structure of the labor market (Danziger and Gottschalk 1995,
Boundetal1995,Juhn1992,2003,Chandra2003,BertrandandMullainathan2003).

Over the last decade or so, labor force participation among black men has deteriorated even
further.Oneoftheleadingexplanationsforthisdevelopmentisthesteepriseinincarcerationof
young black males. Freeman (2003) estimated that roughly 20 percent of the black male
population is incarcerated. In addition to the obvious incapacitating effect of current
incarceration,evenpreviousincarcerationcanhaveanegativeeffectonlaborforceactivity.For
one, incarceration tends to weaken employment networks (Travis, Solomon, and Waul 2001).
Moreover,anauditstudyofemployersinMilwaukeebyPager(2003)suggeststhatemployers
seemmuchmoreaversetohiringblackmaleswithcriminalrecordsthancomparablewhitemen,
andthatthisdeterrenceextendseventothosewithoutcriminalrecords(Holzer,Raphael,and
Stoll 2002, 2006). These demandside factors serve to reinforce supplyside factors that keep
blackmenoutofthelabormarket.

Holzer,Offner,andSorenson(2005)testthishypothesisintheiranalysisofblackmalesages16
34 with a high school education or less in the 1980s and 1990s. They find that previous
incarcerationcanaccountforaquarterormoreofthedeclineinparticipationamongblackmales
62
66
70
74
78
82
1972 1980 1988 1996 2004 2012
Figure22:MaleLaborForceParticipation
RatesbyRace
2014:Q2
Percent
White
Black
39

ages1624andatleasthalfofthedeclineamongblackmalesages2534.Theseestimatesofthe
effectsofincarcerationarecorroboratedbyworkbyWesternandPettit(2000).

40

V. The Outlook for the Participation Rate

Assuming a continuation of current policies, over the longer term the participation rate will
continue to trend down as the aging of the population continues. Over the next few years,
however,itislikelythatthecontinuedrecoveryoftheeconomywilltendtopushintheother
directionfromtheagingtrend.Theevolutionoftheresidualcomponentofparticipationisharder
to predict given the uncertainty about its exact cause. All told, over the next few years, the
participationratecouldriseorfallslightlyfollowedbyaresumptionofitslongertermdecline
duetothecontinuedagingofthepopulation.Policyhastheabilitytoaffectthesetrendsaswe
discussinSectionVI.

WhilethefirstbabyboomersbecameeligibleforearlySocialSecuritybenefitsin2008,thelast
babyboomerswillnotreachthispointoftheirlivesuntil2026.Theshareofthepopulationover
theageof55willcontinuetogrowrapidlyduringthecomingdecade,andmostofthisgrowth
will be in the share of the population over the age of 65 (Figure 23). While the share of the
populationoverage65hasrisengraduallyinthepastduetoincreasedlongevity,theretirement
ofbabyboomersacceleratedthisgrowthanditisnowrisingfasterthanatanypointsince1960.
Iftheparticipationprofilebyagewerefixedat2013levels,thechangingpopulationsharewould
subtractabout0.2percentagepointperyearfromtheparticipationratethrough2024.

AstheeconomyhasrecoveredfromtheGreatRecession,thecyclicaldragontheparticipation
ratehaslessenedtoabout0.5percentagepointinthesecondquarterof2014.Astheeconomy
continuestoreturntofullemployment,thiscyclicalfactorwillcontinuetodissipate.Ifthisgap
weretobeclosedoverabouttwoyears,thenthedissolutionofthecyclicaldragonparticipation
wouldaddabout0.2percentagepointperyeartotheparticipationrateforthenexttwoyears
roughlyoffsettingtheagingtrend.

41

The residual component of the decline in participation represents the greatest source of
uncertaintyabouttheparticipationrategoingforward,inpartreflectingtheuncertaintyabout
its causes. This component may shrink to zero, reflecting temporary factors dissipating as the
economyrecovers.Ontheotherhand,tothedegreeitrepresentstrendsthatpredatetheGreat
Recession,thenabsentpolicyshiftsorotherunexpecteddevelopmentsitcouldpersistoreven
growlargerAaronsonet.al(2006)projectedthatitwouldreducetheparticipationrateby0.2
percentagepointperyearaboveandbeyondthenormalagingtrend.Moreover,whilelongterm
unemployment and other indicators of labor market challenges have been steadily improving,
thereisalwaysadangerthatwithoutpolicychangessomeindividualswouldpermanentlydrop
out of the labor force, resulting in a participation rate that would be permanently lower as a
result. It also remains possible that the residual could grow over the next few years, as it has
grownconsiderablyoverthepastfewyearsandmaycontinuetodosointhenearfuture.

Ultimately, absent changes in policies, a meaningful increase in the participation rate from
currentratesappearsunlikely.Figure24showsseveralscenariosfortheparticipationrateunder
current policies, with a common assumption about the aging trend but varying assumptions
abouttheevolutionofthecyclicalandresidualcomponentsdiscussedabove.

Thesescenariosare
notprojectionsanddonotrepresentthefullrangeofpossibilities,buttheydoindicatethatthe
participationrateislikelytoberoughlystableintheneartermandwilldecreasemorerapidly
aftertheeconomyreturnstofullemploymentandtheagingeffectsonceagaindominate.

42

VI. The Presidents Agenda and the Labor Force Participation Rate

The outlook for the participation rate, however, ultimately depends, at least in part, on policy
choicesthatwemakeandthepracticesadoptedbycompaniesandindividuals.Wediscusssome
ofthemajorpoliciestheAdministrationispursuingtoimprovethetrajectoryoftheparticipation
ratehere.

Aging of the Population


Itisimportanttoputtheagingtrendinperspective.Thatolderworkersareabletoretireisin
many ways a positive development. But it also creates challenges, especially for overall fiscal
policyandinparticularforprogramslikeSocialSecurityandMedicare.

Partoftheanswertothesechallengesisenablingolderworkerswhowanttocontinuetowork
todoso.Workplaceflexibilitypoliciesthatallowemployeestoworkparttimeorflexiblehours
canhelpolderworkerstailortheirjobstotheirspecificneedsandcancreaterolesthatareless
demandingthanfulltimeemployment.InthisyearsStateoftheUnion,thePresidentcalledon
employers to expand access to flexible workplace policies. For similar reasons, many older
workerschooseselfemployment,optingtoretirepartiallyorsettheirownhours.Manyother
older workers would like to be selfemployed, but have medical conditions that make health
insuranceanecessity.Althoughinthepasttheseworkerswouldhavehadtoseekworkatalarge
employerofferinghealthinsurancecoverage,theAffordableCareActenablestheseworkersto
becomeselfemployedbyprovidingaccesstoquality,affordablehealthinsurance(Fairlie,Kapur,
andGates2011)untiltheyareeligibleforMedicare.

Probably the most significant policy response to falling labor force participation rates is
immigrationreform,whichwouldcounteractthelaborforceeffectsofanagingpopulationand
spureconomicgrowth.Onaverage,immigrantsareyoungerandparticipateinthelaborforceat
higherratesthannativebornAmericans.TheCongressionalBudgetOffice(2013)estimatesthat
the immigration reform bill passed by the Senate would increase the labor force by 6 million
people,orfivepercent,by2023,andraisethelaborforceparticipationrateby0.7percentage
point.Thistrendwouldcontinueovertime,providingarangeofothereconomicbenefitssuch
asincreasingGDP,loweringbudgetdeficits,andimprovingSocialSecuritysolvency.

Trends Within Demographic Groups


Policies can also help address the longterm trends in participation within age groups. The
participationratesforprimeagemenandwomenhavedeclinedsteadilyoverthepastdecade,
continuingadecadeslongtrendamongmenandreversingadecadeslongtrendamongwomen.
Among younger workers, participation rates have fallen more rapidly. The fall in participation
amongyoungblackmenhasbeenparticularlydrastic.ThePresidentsagendaaddressesthese
trends and lays out a series of policies that could help encourage individuals to remain in the
laborforce.

43

To encourage more women to participate in the labor force, the Administration has a robust
policy agenda around working families, which was further developed at the White House
Working Families Summit in June 2014. The summit brought together government, private
sector,labor,andacademicleaderstodiscusshowtohelpworkersbalancetheirpersonaland
professional lives. Policies such as paid family leave, flexible work schedules, and affordable
childcare are vital for families, and mothers in particular. Such policies not only encourage
participationinthelaborforce,buttheyalsohelppeoplefindthejobsbestsuitedtotheirskills
andexcelinthosepositions.Giventhatwomenoftentakeonwelloverhalfoffamilyobligations,
suchpoliciesareparticularlyhelpfulinraisingwomenslaborforceparticipation(BlauandKahn
2013).Aswomennowmakeupnearlyhalfofthelaborforceandearn59percentofallhigher
educationdegrees,policiesthatfacilitatetheirparticipationinthelaborforcecanhavealarge
impactonthelaborforceparticipationrateandoveralleconomicgrowth.

Another very successful policy tool to encourage labor force participation among lowincome
workershasbeentoraisethereturnstoworkthroughtheEITC,withthe1980sand1990sEITC
expansionspullingmorethanhalfamillionpeopleintothelaborforce.ButtheEITCavailableto
workerswithoutchildrenandnoncustodialparents(thechildlessworkerEITC)isverysmall,
andthecurrentrulespreventworkersyoungerthan25andolderthan64fromclaimingthecredit
atall.ThePresidenthasproposedtodoubletheEITCavailabletoworkerswithoutchildrenand
noncustodialparentsandtomakeitavailabletoyoungerandolderworkers.Thispolicywould
targetanumberofgroupswithfallingorlowlaborforceparticipationrates,includingminority
men,youngadultsnotenrolledinschool,workerswithdisabilities,andolderworkers.

Inaddition,tospecificallysupportyoungmenofcolor,thePresidentlaunchedtheMyBrothers
Keeper Initiative in February 2014. This initiative created a Federal Task Force charged with
assessing which public and private efforts are most effective at helping young men of color
succeed academically, in their communities, and in the workplace. So far, eleven leading
philanthropic foundations have committed to investing $200 million to study which programs
keeptheseyoungmenontrackandbringtheseprogramstoscale.

Finally, building labor force attachment among young people can raise their labor force
participation throughout life and offset the longterm declining trend. Earlier this year, the
Presidentannounced$100millioninfundingforAmericanApprenticeshipGrants,whichwilluse
existing funds to expand apprenticeships, training young people for highdemand jobs, and
helping them build relationships with employers. Additionally, the Administration will award
$500milliontocommunitycollegestoimplementjobtrainingprogramsbasedontheskillsand
credentialsindemandamongbusinesses.ThePresidenthassimilarlycommittedtoredesigning
highschoolcurriculatofocusmoreonindustryrelevantskills.Alloftheseinitiativesaimtogive
youngpeoplelabormarketskillsandconnectthemwithemploymentopportunities,raisingtheir
laborforceparticipationoverthelongterm.

44

The Cyclical Component of Participation


While the sections above highlight policies that can alter longterm labor force participation
trends,anumberofpoliciescouldnoticeablyreducethecyclicaldragontheparticipationrate
by strengthening our economic recovery and increasing aggregate demand. For example,
increasinginvestmentininfrastructurewouldprovideanimmediateboosttoemploymentand
demand, as well as the longterm benefits of improved transportation systems. The President
recently proposed a $302 billion fouryear transportation reauthorization plan that would
supporthundredsofthousandsofjobs.

Likewise,thePresidentsproposaltoincreaseinvestmentsinresearch,education,infrastructure,
security, and other critical areas through the Budgets Opportunity, Growth, and Security
Initiative would both speed the return to full employment and help lay the groundwork for
strongerlongrungrowth.Bystrengtheningtheeconomy,thisproposalcouldraisethecyclical
portionofthelaborforceparticipationrate.

ExtendingEmergencyUnemploymentCompensation(EUC)wouldalso boosteconomicactivity
and aggregate demand by supporting jobseekers incomes while they are unemployed.
Moreover,researchshowsthatEUChelpslongtermunemployedworkersstayconnectedtothe
laborforceandkeeplookingforwork.

Otherstepswouldalsohelp,likereauthorizingtheExportImportBank.Intheabsenceofaction
fromCongress,PresidentObamawillpressforwardbyusinghisexecutiveauthoritytospeed
thepermittingofinfrastructureprojects,launchnewhubsofmanufacturinginnovation,and
attractforeigninvestment.

TheAdministrationisalsotakingactiononitsowntostrengthentherecoveryofthehousing
market.InMay,theFederalHousingAdministration(FHA)embarkedonaprogramtoaddress
concernsaboutaccessibilityofmortgagecredit,andactingindependently,FederalHousing
FinanceAgency(FHFA)DirectorMelWatthasalsoundertakenstepstoprovideclarityonso
calledputbackriskandtherebyeasecreditconstraintsinthemortgagemarket.Also,the
AdministrationhasextendedtheMakingHomeAffordableprogramtocontinuetohelp
strugglinghomeownersfacingforeclosureorwhosemortgagesareunderwater.

Factors Unique to the Great Recession: Longterm Unemployment


AcompleterecoveryfromtheGreatRecessionwillalsorequireaddressingtheproblemoflong
term unemployment. Policies are needed to help the longterm unemployed find jobs and
preventthemfromleavingthelaborforce;thesamepoliciesmayalsohelpencouragesomeof
those who have left the labor force over the last few years to reenter it. One issue that
researchers have identified is discrimination against the longterm unemployed by employers
whooftenusecomputerprogramsorpoliciestoscreenoutthelongtermunemployedfromjob
applicants.Longtermunemployedjobseekershavetosendout3.5timesasmanyapplications
asshorttermunemployedseekerstogetaninterview(Ghayad2013).Toaddressthisandother
issuesfacingthelongtermunemployed,thePresidentconvenedprivatesectorleaderstodiscuss
45

howthegovernmentandprivateindustrycanworktogetheronthischallenge.Morethan300
CEOssignedontobestpracticesforrecruitingandhiringthelongtermunemployed,including
80 of the nations largest businesses and over 20 from the Fortune 50. The Administration is
continuing to work with employers to learn and build on what works in helping the longterm
unemployedreturntowork.

To further this cause, the Administration will use $150 million in existing funds to scale up
partnershipsbetweenemployersandnonprofitsthathelpgetthelongtermunemployedback
to work through workbased training, job placement assistance, and employer outreach. The
Administration also recognizes the value of extending EUC benefits in keeping the longterm
unemployed workers attached to the labor force. Bringing down the rate of longterm
unemploymentthroughpoliciessuchasthese,andstrengtheningtherecoveryoverall,mayhelp
resolvetheresidualeffectonlaborforceparticipation.

46

References

Aaronson,Daniel,JonathanDavis,andLuojiaHu.2012.ExplainingtheDeclineintheU.S.Labor
ForceParticipationRate.ChicagoFedLetter296.
Aaronson,Daniel,KyungHongPark,andDanielSullivan.2006.TheDeclineinTeenLaborForce
Participation.EconomicPerspectivesQ1:218.
Aaronson,Stephanie,BruceFallick,AndrewFigura,JonathanPingle,andWilliamWascher.
2006.TheRecentDeclineintheLaborForceParticipationRateandItsImplicationsfor
PotentialLaborSupply.BrookingsPapersonEconomicActivitySpring2006:69154.
Abe,Masahiro,YoshioHiguchi,andJaneWaldfogel.1998.MaternityLeavePoliciesand
WomensEmploymentafterChildbirth:EvidencefromtheUnitedStates,Britainand
Japan.CentreforAnalysisofSocialExclusion,LSE.
Autor,David,H.,DavidDorn,andGordonH.Hanson.2013.UntanglingTradeandTechnology:
EvidencefromLocalLaborMarkets.WorkingPaper18938.Cambridge,MA:National
BureauofEconomicResearch.
Autor,David,andDuggan,Mark.2003.TheRiseintheDisabilityRollsandtheDeclinein
Unemployment.QuarterlyJournalofEconomics,118(1):157206.
Autor,DavidH.,LawrenceF.Katz,andMelissaS.Kearney.2008.TrendsinU.S.Wage
Inequality:RevisingtheRevisionists.ReviewofEconomicsandStatistics.
Bengali,Leila,MaryDaly,andRobValletta.2013.WillLaborForceParticipationBounceBack?
FederalReserveBankofSanFranciscoEconomicLetter,May13,2013.
Bertrand,Marianne,andSendhilMullainathan.2003.AreEmilyandGregMoreEmployable
thanLakishaandJamal?AFieldExperimentonLaborMarketDiscriminationWorking
Paper9873.Cambridge,MA:NationalBureauofEconomicResearch.
Blau,DavidM.,andRyanM.Goodstein.2010.CanSocialSecurityExplainTrendsinLabor
ForceParticipationofOlderMenintheUnitedStates?JournalofHumanResources45
(2):32863.
Blau,Francine,D.,andLawrence,M.Kahn.2013.FemaleLaborSupply:WhyistheUSFalling
Behind?WorkingPaper7140.IZADiscussionPapers.
Borjas,GeorgeJ.,JeffreyGrogger,andGordonH.Hanson.2008.Immigrationandthe
EconomicStatusofAfricanAmericanMen.Economica,77:255282.
Bound,JohnandRichardFreeman.1992.WhatWentWrong?TheErosionofRelativeEarnings
andEmploymentamongYoungBlackMeninthe1980s.QuarterlyJournalofEconomics
108:20132.
Bound,John,MichaelSchauenbaum,andTimothyWaidmann.1995.RaceandEducation
DifferencesinDisabilityStatusandLaborForceAttachmentintheHealthand
RetirementSurvey,JournalofHumanResources30(Supplement1995):S227S267.
(BLS)BureauofLaborStatistics.EmploymentProjections201020.USDepartmentofLabor.
February1,2012.
Casselman,Ben.Missing:UpTo4MillionWorkers.Fivethirtyeight.com.March26,2014.
Chandra,Amitabh.2003.IstheConvergenceoftheRacialWageGapIllusory?WorkingPaper
9476.Cambridge,MA:NationalBureauofEconomicResearch.
47

Coile,Courtney,C.2004.RetirementIncentivesandCouplesRetirementDecisions.Topicsin
EconomicAnalysisandPolicy,4(1).
(CBO)CongressionalBudgetOffice2013.TheEconomicImpactofS.744,theBorderSecurity,
EconomicOpportunity,andImmigrationModernizationAct.CongressionalBudget
Office.June2013.
.2014.TheSlowRecoveryoftheLaborMarketCongressionalBudgetOffice.February2014.
Cowen,Tyler,andJaymeLemke.2011.10PercentUnemploymentForever?ForeignPolicy.
January5,2011.
Dahl,GordonB.,KatrineV.Lken,MagneMogstad,andKariVeaSalvanes.2013.WhatIsthe
CaseforPaidMaternityLeave?WorkingPaper19595.Cambridge,MA:NationalBureau
ofEconomicResearch.
Danziger,SheldonandGottschalk,Peter.1995.AmericaUnequal.NewYork:RussellSage
Foundation.
Elsby,Michael,BartHobijn,andAyseglSahin.2013.OntheImportanceoftheParticipation
MarginforLaborMarketFluctuations.Unpublishedmanuscript.
Erceg,ChristopherJ.,andAndrewT.Levin.2013.LaborForceParticipationandMonetary
PolicyintheWakeoftheGreatRecession.FederalReserveBankofBoston,April2013.
Fallick,BruceandJonathanPingle.2007.ACohortBasedModelofLaborForceParticipation.
BoardofGovernorsoftheFederalReserveSystem,FinanceandEconomicsDiscussion
SeriesWorkingPaper200709.
.2013.DataupdatetoACohortBasedModelofLaborForceParticipation.September,
2013.
Fairlie,RobertW.,KanikaKapur,andSusanGates.2011.IsEmployerbasedHealthInsurancea
BarriertoEntrepreneurship?JournalofHealthEconomics,30(1):146162.
Fernandez,Raquel.2013.CulturalChangeasLearning:TheEvolutionofFemaleLaborForce
ParticipationoveraCenturyAmericanEconomicReview,103(1),472500.
Fogg,NeetaP.andPaulE.Harrington.2011.TheCollapseoftheLaborMarketfor1624Year
Olds.Cascade:No.78.Fall2011.
Freeman,Richard,B.2003.CanWeClosetheRevolvingDoor?:RecidivismvsEmploymentof
ExOffendersintheU.S.UrbanInstituteReentryRoundtable.May1920,2003.
Fujita,Shigeru.2014.OntheCausesofDeclinesintheLaborForceParticipationRate.Federal
ReserveBankofPhiladelphia.February6,2014.
Ghayad,Rand.2013.TheJoblessTrap.WorkingPaper,2013.
Goldin,ClaudiaandLawrenceF.Katz.1998.TheOriginsofTechnologySkillComplementarity.
TheQuarterlyJournalofEconomics113(3):693732.
.2007.LongrunchangesintheUSwagestructure:narrowing,widening,polarizing.
WorkingPaper13568.Cambridge,MA:NationalBureauofEconomicResearch.
Gustman,AlanL.andThomasSteinmeier.2008.HowChangesinSocialSecurityAffectRecent
RetirementTrends.WorkingPaper14105.Cambridge,MA:NationalBureauof
EconomicResearch.
Gustman,Alan,L.,Thomas,L.Steinmeier,andNahidTabatabai.2011.HowDidtheRecession
of20072009AffecttheWealthandRetirementoftheNearRetirementAgePopulation
intheHealthandRetirementStudy?WorkingPaper17547.Cambridge,MA:National
BureauofEconomicResearch.
48

Holzer,HarryJ.,PaulOffner,andElaineSorrenson.2005.DecliningEmploymentAmongYoung
BlackLessEducatedMen:TheRoleofIncarcerationandChildSupport.JournalofPolicy
AnalysisandManagement,24(2),329350.
Holzer,HarryJ.,StevenRaphael,andMichaelA.Stoll.2006.PerceivedCriminality,Criminal
BackgroundChecks,andtheRacialHiringPracticesofEmployers.JournalofLawand
Economics49(2)451480.
.2002.HowdoCrimeandIncarcerationAffecttheEmploymentProspectsofLesseducated
YoungBlackMen?ExtendingOpportunityConference.WashingtonDC:May2002.
Juhn,Chinhui.1992.TheDeclineinMaleLaborMarketParticipation:TheRoleofDeclining
MarketOpportunities,QuarterlyJournalofEconomics107(1):79121.
.2003.LaborMarketDropoutsandTrendsintheWagesofBlackandWhiteMen,Industrial
andLaborRelationsReview56:643662.
Kahn,LisaB.2010.TheLongtermLaborMarketConsequencesofGraduatingfromCollegeina
BadEconomy.LabourEconomics17(2):303316.
Karoly,LynnA.,andJulieZissimopoulos.2004SelfEmploymentamongOlderU.S.Workers.
MonthlyLaborReview,27(7):2447.
Kroft,Kory,FabianLange,andMatthewJ.Notowidigdo.2013.DurationDependenceand
LaborMarketConditions:EvidencefromaFieldExperiment.TheQuarterlyJournalof
Economics.128(3):11231167.
Kroft,Kory,FabianLange,andMatthewJ.NotowidigdoandLawrenceF.Katz.2014.LongTerm
UnemploymentandtheGreatRecession:TheRoleofComposition,Duration
Dependence,andNonParticipation.WorkingPaper20273.Cambridge,MA:National
BureauofEconomicResearch.
Krueger,AlanB.,JuddCramer,andDavidCho.2014.AretheLongTermUnemployedonthe
MarginsoftheLaborMarket?WorkingPaper.Washington,DC:BrookingsPanelon
EconomicActivity.March10,2014.
Kudlyak,Marianna.2013.ACohortModelofLaborForceParticipation.EconomicQuarterly
99(1):2543.FirstQuarter2013.
Maestas,Nicole,andJulieZissimopoulos.2010.HowLongerWorkLivesEasetheCrunchof
PopulationAging.JournalofEconomicPerspectives24(1):139160.
Maestas,Nicole.2007.BacktoWork:ExpectationsandRealizationsofWorkafter
Retirement.JournalofHumanResources45(3):718748.
Mitchell,Josh.2013.Whoarethelongtermunemployed?Washington,DC:TheUrban
Institute.July2013.
Moynihan,DanielP.1965.Employment,Income,andtheOrdealoftheNegroFamily.
Daedalus9(4),TheNegroAmerican.Fall1965.
Pager,Deyah.2003.TheMarkofaCriminalRecord.AmericanJournalofSociology,108(5),
937975.
Pattison,David,andHilaryWaldron.2013.GrowthinNewDisabledWorkerEntitlements,
19702008.SocialSecurityBulletin.73(4).
Reed,Deborah,andSheldonDanziger.2007.TheEffectsofRecentImmigrationon
Racial/EthnicLaborMarketDifferentials.AmericanEconomicReview97(2):37377.
RobinsonP.M.1988.RootNconsistentSemiparametricRegression.Econometrica56:931
954.
49

RossinSlater,Maya,ChristopherJ.Ruhm,andJaneWaldfogel.2013.TheEffectsofCalifornia's
PaidFamilyLeaveProgramonMothersLeaveTakingandSubsequentLaborMarket
Outcomes,JournalofPolicyAnalysisandManagement,32(2),224245,
Schirle,Tammy.2008.WhyHavetheLaborForceParticipationRatesofOlderMenIncreased
sincetheMid1990s?JournalofLaborEconomics26(4):549594.
Shierholz,Heidi.2012.LaborForceParticipation:CyclicalversusStructuralChangessincethe
StartoftheGreatRecession.EconomicPolicyInstitute.IssueBrief#333.May24,2012.
Smith,Christopher.2008.Dude,WheresMyJob?TheImpactofImmigrationontheYouth
LaborMarket.UnpublishedPh.D.dissertation,DepartmentofEconomics,
MassachusettsInstituteofTechnology.
.2011.Polarization,Immigration,Education:What'sBehindtheDramaticDeclineinYouth
Employment?DivisionofResearchandStatisticsandMonetaryAffairs,FederalReserve
Board.
Stock,JamesH.1989.NonparametricPolicyAnalysis.JournaloftheAmericanStatistical
Association84(406):567575.
Travis,J.,AmyL.Solomon,andMichelleWaul.2001.FromPrisontoHome:TheDimensionsand
ConsequencesofPrisonerReentry.WashingtonDC:UrbanInstitutePress.
VanZandweghe,Willem.2012.InterpretingtheRecentDeclineinLaborForceParticipation,
FederalReserveBankofKansasCityEconomicReview,FirstQuarter2012,534.
Western,BruceandBeckyPettit.2000.IncarcerationandRacialInequalityinMens
Employment.IndustrialandLaborRelationsReview,54(1):316.
50

Sources

Figure1:BureauofLaborStatistics.
Figure2:BureauofLaborStatistics.
Figure3:CensusBureau;CEAcalculations.
Figure4:BureauofLaborStatistics;CEAcalculations.
Figure5:BureauofLaborStatistics;CEAcalculations.
Figure6:NationalCenterforHealthStatistics.
Figure7:SocialSecurityAdministration.
Figure8:BureauofLaborStatistics;SocialSecurityAdministration;CEAcalculations.
Figure9:BureauofLaborStatistics;CEAcalculations.
Figure10:BureauofLaborStatistics;CEAcalculations.
Figure11:BureauofLaborStatistics;CEAcalculations.
Figure12:BureauofLaborStatistics;CEAcalculations.
Figure13:BureauofLaborStatistics;CEAcalculations.
Figure14:BureauofLaborStatistics;CEAcalculations.
Figure15:SocialSecurityAdministration;CEAcalculations.
Figure16:BureauofLaborStatistics.
Figure17:NationalCenterforEducationStatistics.
Figure18:BureauofLaborStatistics.
Figure19:BureauofLaborStatistics.
Figure20:InternationalLabourOrganization.
Figure21:BureauofLaborStatistics.
Figure22:BureauofLaborStatistics.
Figure23:CensusBureau;CEAcalculations.
Figure24:BureauofLaborStatistics;SocialSecurityAdministration;CEAcalculations.
Table1:VariousSources;CEAcalculations.
Table2:BureauofLaborStatistics;CEAcalculations.
Table3:BureauofLaborStatistics;CEAcalculations.
Table4:BureauofLaborStatistics;SocialSecurityAdministration;CEAcalculations.
Table5:BureauofLaborStatistics;SocialSecurityAdministration;CEAcalculations.
Table6:BureauofLaborStatistics;CEAcalculations.
AppendixFigure1:BureauofLaborStatistics;CEAcalculations.
AppendixFigure2:BureauofLaborStatistics;CEAcalculations.
AppendixFigure3:BureauofLaborStatistics;CEAcalculations.
AppendixFigure4:BureauofLaborStatistics;CEAcalculations.

51

Appendix A Time Series Analysis Specifications

Construction of a CyclicallyAdjusted Trend Component


A cyclicallyadjusted trend component of the participation rate was estimated using a
semiparametricprocedureforapartiallylinearmodel(Robinson1988,Stock1989),combininga
nonparametricbiweightfilterwithaparametriccyclicaladjustmentstepusingleadsandlagsof
theunemploymentgap.Thebiweightfilter(usingawindowwidthof40quarters)wasappliedto
the first difference of the participation rate and the trend component was constructed by
cumulating the trend in first differences. First differences in the detrended participation were
thenregressedontwoleads,twolags,andthepresentvalueoftheunemploymentgap.
The predicted values from this regression were cumulated to produce the estimated cyclical
component, which was subtracted from the participation rate to adjust for the business cycle.
Thecyclicallyadjustedserieswasthenrunthroughthebiweightfilterinfirstdifferencesagain
and the estimates were cumulated to produce the estimated trend component of the
participationrate(AppendixFigure3).
Lookingattheperiodsincetheendof2007,theestimatedtimeseriestrendwasverysimilarto
theestimatedagingeffect,indicatingthatchangesinparticipationwithinagegroupsmayhave
offsetandproducedonlyasmallreductionintotalparticipation.


52




53


62
63
64
65
66
67
68
1980 1990 2000 2010
(1) (5)
(4) (8)
AppendixFigure4:EstimatedCyclicalComponents
from Regressions1,4,5,and8,19802014
2014:Q2
Percent
Actual

Вам также может понравиться