Communication
Are Children Worse Off'?
EvaluatingWell-Being Using a New
(and Improved)Measure of Poverty
John Iceland
Kathleen Short
Thesia I. Garner
David Johnson
ABSTRACT
Although child poverty rates continue to surpass those of others, there is
growing consensus that current official poverty measure has become outdated and flawed. Using data from the Current Population Survey and the
Survey of Income and Program Participation, we implement an experimental poverty measure based on recommendations by a National Academy of
Sciences panel. We find that while child poverty rates continue to surpass
those of others, the gap between child and adult poverty rates is smaller
under the experimental measure. Results highlight the impact of noncash
government benefits and the Earned Income Tax Credit in reducing child
poverty.
John Icelandand KathleenShortare researchersat the U.S. CensusBureau.ThesiaL.Garnerand David Johnsonare researcheconomistsat the Bureauof LaborStatistics.Public-useversionsof the survey data analyzed here are available from the U.S. Census Bureau, Direct all correspondence and
questions about the analysis to John Iceland, HHES Division, Building 3, Room 1472, U.S. Census Bureau, Washington, D.C. 20233-8500, [email protected]. This paper reports the results of research
and analysis undertaken by Census Bureau and Bureau of Labor Statistics staff. It has undergone a
more limited review than official Census Bureau publications. This report is released to inform interested parties of research and to encourage discussion. The authors would like to thank Jeff Sisson and
anonymous reviewers for their contributions to this article.
[Submitted May 1999; accepted July 2000]
THE JOURNAL
OF HUMAN
RESOURCES
* XXXVI
* 2
Communications:
Iceland,Short,Garner,and Johnson 399
I. Introduction
Thereis growingconsensusthat the way povertyis currentlymeasuredin the UnitedStatesis outdatedand flawed(CitroandMichael 1995;Ruggles
1990).The currentofficialpovertymeasure,originallyadoptedin the 1960s,consists
of a set of thresholdsfor families of differentsizes and compositionthat are comparedto a familyresourcemeasureto determinea family'spovertystatus.Basically,
the thresholdsrepresentthe cost of a minimumdiet multipliedby threeto allow for
expenditureson othergoods and services,and familyresourcesaredefinedin terms
of gross cash income.
Accordingto the official measure,trendsin child povertyare not encouraging.
After a periodof improvementin the 1960s, child povertyworsenedover the last
threedecades.In 1997, 19.9 percentof childrenin the U.S. were poor (Dalakerand
Naifeh 1998). In contrast,in 1997 the povertyratesfor people aged 18-64 and the
elderly were 10.9 and 10.5 percent,respectively.As of 1997, childrenconstituted
about40 percentof the povertypopulation,thoughonly abouta quarterof the total
population.
The officialpovertyratemasksthe full extentof the impactof governmentbenefits
on povertyreductionbecauseit only countscash income in the measureof family
resources,while the growthin means-testedtransfersin recent decades has been
overwhelminglyconcentratedin noncashandtax expenditureprograms,suchas food
stampsand the EarnedIncome Tax Credit(EITC)(Centeron Budget and Policy
Priorities1998). Becausemanyof these transfersaredesignedto help familieswith
children,the currentofficialpovertymeasuremay overstatepovertyamongchildren
relative to poverty among other demographicgroups (Betson and Warlick 1998;
Citroand Michael 1995; Shortet al. 1999).
Concernsaboutthe adequacyof the povertymeasureincreasedduringthe past
two decades,culminatingin a Congressionalappropriation
for an independentscientific study of povertymeasurementissues. In response,the NationalAcademy of
Sciences (NAS) establishedthe Panelon Povertyand FamilyAssistance,which released its report titled Measuring Poverty: A New Approach in 1995 (Citro and Mi-
chael 1995). The NAS panel recommendeda new measurethatbetterreflectscontemporarysocial and economicrealitiesand governmentpolicy.
The NAS panelidentifiedseveralmajorweaknessesin boththe thresholdandthe
resourcedefinitionof the currentmeasure.The two most importantones relevantto
our understandingof child povertyare: (1) The currentincome measuredoes not
reflectthe effects of key governmentpolicies thatalterthe disposableincomeavailable to families;and (2) The currentmeasuredoes not take into accountvariation
in expensesthatare necessaryto hold a job and to earnincome, such as child-care
costs andtaxes.Becauseof these andotherdeficiencies,the currentpovertymeasure
does not accuratelyreflect the economic well-being of many people (Citro and
Michael 1995).
FollowingtheNAS Panel'srecommendations,
theCensusBureaureleaseda report
in 1999 with severalexperimentalpovertymeasures(Shortet al. 1999). This report
containeda few summarystatisticson childrenindicatingthat the currentofficial
measuremay indeed overestimatepovertyamongchildren.
This paperbuildson the NAS Panel'sworkand the subsequentCensusreportby
400
The Journalof HumanResources
focusingmore closely on the implicationsof moving to a new measureof poverty
of child poverty.We do this by exploringthe elementsof the
on our understanding
new povertymeasurethataffectchildpovertyestimatesin particularandcalculating
povertyratesusing both the CurrentPopulationSurvey (CPS), the currentsource
of official poverty statistics, and the Survey of Income and ProgramParticipation (SIPP), the recommendedsource of futurepoverty data (Citro and Michael
1995).
II. Data and Methods
We estimatepovertyrates for 1992-96 using data from the March
supplementto the CurrentPopulationSurvey(CPS), and 1992 povertyrates from
(SIPP).A primary
the 1992panelof the Surveyof IncomeandProgramParticipation
recommendationof the NAS panel was to make the SIPP the principalsource of
povertystatisticsbecauseit asks more relevantquestionsand obtainsincome data
of higherqualitythan the CPS, but more researchand developmentis needed on
the SIPP before it can supplantthe CPS in this role. Perhapsmost important,the
SIPPstill lacks reliabletax estimates(a tax model is currentlyunderdevelopment),
which is particularlyimportantfor examiningthe growingimpactof the EITCon
experimentalpovertymeasures.In addition,moretimelyreleaseof SIPPdatais also
necessary.
Ouranalysisproceedsas follows. First,resultsfrom both the SIPPand CPS are
presentedcomparinghow the officialchild,adult,andelderlypovertyratesfor 1992
are affectedby implementingvariousNAS panel recommendations(listed below)
one at a time, highlightingthe effect of governmentbenefitsin particular.Second,
resultsarepresentedshowingpovertyratesfor 1992 when all the recommendations
are simultaneouslyimplemented.We thendiscussthe distributionof povertyamong
subgroupsof childrenin thatyear and trendsin povertybetween 1992 and 1996.
Underthe experimentalpovertymeasure,a family's resourcesare definedas the
valueof moneyfromall sourcesplus the valueof near-moneybenefitsthatareavailableto buy goods andservicescoveredby the new thresholds,minusnon discretionaryexpenses.Near-moneybenefitsor tax refundsthatarenot countedin the official
definitionof income includethe following:food stamps,housingsubsidies,school
lunch subsidies,home energy assistance,and the EarnedIncome Tax Credit.Expenses subtractedinclude:income and payrolltaxes (includingcapitalgains/losses
estimates),child-careandotherwork-relatedexpenses,andhouseholdcontributions
towardthe costs of medical care and healthinsurancepremiums(medicalout-ofpocket costs).
In our SIPPpovertyestimates,we also subtractfrom income child supportpayments made by the payer,and add to income the value of benefitsreceivedunder
the Women,Infants,andChildrennutritionprogram(WIC)andthe schoolbreakfast
program.Estimatesof these three elements are not availablein the CPS. In our
povertyestimatesusing SIPP data,we do not includethe effect of state taxes and
our SIPP
capital/gainsbecausereliableestimatesarenot yet available.Furthermore,
Communications:
Iceland,Short,Garner,and Johnson 401
federalincome tax, EarnedIncomeTax Credit,and payrolltax estimatesare only
roughestimateswhich shouldbe viewed with caution.)
Povertythresholdsunderthe experimentalpovertymeasureare representedby a
dollaramountfor food, clothing,shelter,utilities(FCSU),as well as a smallamount
to allow for otherneeds (such as householdsupplies,personalcare).2As the panel
recommended,a thresholdis developedfor a referencefamily type consisting of
two adultsandtwo childrenusing consumerexpendituresurveydata,andthe reference family thresholdis then adjusted,using an equivalencescale, to reflect the
needs of differentfamily types. We use a three-parameter
equivalencescale in this
analysis.3Furtheradjustmentsare madeto reflectgeographicdifferencesin housing
costs.4Finally,thresholdsareupdatedovertimebasedon estimatesof medianexpenditureson FSCU items in the ConsumerExpenditureSurvey.
This analysis basically uses an experimentalpoverty measureidentical to the
"DES-DCM2"measurecontainedin the Census Bureaureport,which containsa
different,andarguablymorerefined,equivalencescaleandmethodfor valuingchildcare costs thanrecommendedby the NAS panel. See Shortet al. (1999) for more
detailson the constructionof each of the elementsof this experimentalpovertymeasure.
It is importantto note thatresearchon severalelementsof the experimentalpovertymeasures-including geographicadjustmentsto thresholds,equivalencescales,
child-careexpenses, and medical out-of-pocketexpenses-continues. Perhapsthe
element generatingthe most controversyis the estimateof medical-out-of-pocket
expenses.5In general, subtractingmedical-out-of-pocketexpenses increases esti1. SIPP payrolltaxes were imputedbased on self-employmentand wage and salaryincome, similarto
the standardmethodused in the CPS. Federalincometaxes and the EITCwere calculatedbasedon the
assumptionthatall familiestook the standarddeduction.The resultingestimatesarethereforeonly rough
of total tax liabilities,particularlyamonghigher-incomefamilies.
approximations
2. The NAS Panel recommendthat thresholdsshouldbe set at betweenthe 30th and 35th percentiles
medianexpenditureson FCSU.In the CensusBureaureport(Shortet al. 1999),as in this analysis,thresholds used are at the midpointof this range.
3. In the threeparameterscale,proposedby Betson(1996), one parametertakesinto accountthatchildren
consumeless thanadults,anotherthatthereare economiesof scale in largerfamilies(as a familyof six
usuallydoes notusuallyspendtwiceas muchon basicneedsas a threepersonfamily),anda thirdparameter
providesmoreeconomiesof scale betweensingles andchildlesscouples,andmoresimilaritybetweenthe
scales for familiesof one parentwith two childrenand two parentswith one child, thanan equivalence
scale with just two parameters.Betson's scale is definedfor each of threedifferentfamily types as: (1)
two-adultonly: 1.41; (2) For single-parentfamilies:(a + 0.8 + p*(c - 1))F; (3) and for otherfamilies:
(a + p*c)F; wherep = 0.5, and F = 0.7, a = numberof adultsin family, and c = numberof children
in family.
4. Followingthe NAS panel,we use interareahousingcost indexescalculatedfrom 1990 Censusdataon
grossrent.TheNAS panelfocusedon sheltercosts in the geographicadjustments
becausehousingexpendituresarethe largestcomponentof thepovertybudgetandbecausevariationsin housingcostsaresignificant
acrossregionsandby populationsize. Theseindexesareproducedfor six populationsize categorieswithin
each of the nine censusregions.For example,the thresholdfor the referenceunit in a largemetropolitan
areain New Englandis 27 percenthigherthanthe nationalaverage,while it is 15 percentlower thanthe
nationalaverageif this family lives in a nonmetropolitan
areain the West SouthCentralregion.
5. Followingthe recommendations
by the NAS Panel,these out-of-pocketexpensesare imputedto both
CPS andSIPPfamiliesin this analysisbasedon datafromthe 1987 NationalMedicalExpenditureSurvey
(NMES).The approachinvolves allocatingmedicalexpendituresto families based on characteristicsof
the family head and calibratingthose allocationsto controltotals.
402
The Journalof HumanResources
mated elderly povertyrates relativeto those of working-ageadults and children.
Althoughthereis generalconsensusthatsuchmedicalexpensesshouldbe takeninto
accountin a new povertymeasure(Citroand Michael 1995), the preciseapproach
is subjectto futuremodification.
Overall,conclusionsregardingchild povertyratesvary only modestlyregardless
of the exact experimentalpoverty measureused (Short et al. 1999). Even if the
methodsof estimatingmedicalexpenses are modifiedin the future,conclusionsin
this paperaboutthe impactof otherelementsin the experimentalpovertymeasure
still hold,as well as the generalfindingsthateven thoughthe currentofficialmeasure
tends to overestimatechild povertyratesrelativeto adultrates,child povertyrates
still surpassthose of others.We now discuss these resultsin more detail.
III. Results
Table 1 displaysdatafromthe 1992 CPS and SIPPon benefitsand
in a refinedmeasureof a family'sdisposableincome.
expensesthatareincorporated
The tableindicatesthat,accordingto the CPS, about11 percentof all personslived
in families that received food stamps.Food stampreceipt is higher among poor
families,andespeciallyamongpoorfamilieswithchildren.Abouttwo thirdsof CPS
children-and morethanthree-quarters
of SIPPchildren-lived in poorfamiliesthat
receivedfood stamps.Food stampincome averagedabout$2,500 in such families
accordingto the CPS.
Subsidizedschool lunchesareanothertransferreceivedby manychildrenin poor
families,with an averagesubsidyof $660 accordingto the CPS. Housingsubsidies,
althoughmore substantial,are receivedby fewer families-about 23 percent,accordingto the CPS, and close to 31 percentin the SIPP.6Overall,resultsshow that
poorchildrenaremorelikely to be in familiesthatreceivevarioustypes of noncash
assistancethanotherpoor people.
In general,the tablealso indicatesthatbenefitsin the SIPPtendto be higherthan
those fromthe CPS. This follows fromthe fact thatthe SIPP,as an income survey,
is designedto do a morecompletejob of collectingincomedata(Coderand ScoonRogers 1996). Respondentrecall tends to be betterunderthe SIPP design, where
peopleareinterviewedeveryfourmonthsratherthanannuallyas in the CPS.Finally,
smalldifferencesin questionwordingandperiodof recallfor the school lunch,food
stamps,and energyassistanceitems in the surveysmay play a role.
Of the expenses listed in the table, work-relatedexpenses and medical out-ofpocketcosts tendto be incurredby a high proportionof most types of families,and
the amountsarequitesubstantial.Child-careexpensesarealso substantial,withmore
6. Differentmethodswere used for estimatinghousingsubsidiesin the SIPP and CPS analyses.In the
CPSpovertymeasure,estimatesarecalculatedusingstatisticalmodelsbasedon thedatafromthe American
HousingSurvey.In the SIPP measure,subsidiesare estimatedbased on county-levelFairMarketRents
(FMR)for housingunits of differentsizes.
Iceland,Short,Garner,and Johnson 403
Communications:
than 10 percentof people living in familiesthatreportsuch expenses,accordingto
the CPS.7Finally, althoughtaxes tend to impose a substantialfinancialburdenon
a majorityof families,poorfamiliespay considerablyless. In fact, manypoorfamilies (abouthalf of poor children'sfamilies) get tax relief in the form of the EITC,
accordingto both CPS and SIPPestimates.
Table2 shows the impactof each resourceadditionor subtractionon the poverty
rates for each of three age groups-children, adults 18-64, and the elderly.These
measuresprovidean incrementalview of the effect of each recommendedchange.
The last two rows of Table 2 also show the completeexperimentalpovertyrates
and how they differ from the official ones.
Lookingat variouselementsthataddto a family's income (noncashgovernment
benefits,the EITC,and capitalgains), we find thatfood stampsand the EITCtend
to have the biggest impact on povertyrates. The marginaleffect is such that the
povertyrate declines from 14.8 percentto 14.1 when each of these two sourcesof
incomeareaddedseparatelyto the officialresourcemeasure,accordingto datafrom
the CPS. Housing subsidies alone have a modest estimatedimpact, lowering the
povertyrate from 14.8 to 14.4 percent.
Thereis some variationby age group.Food stampshave the biggest impacton
child povertyrates(loweringthe child povertyratefrom 22.3 to 21.0, accordingto
CPS data). The EITC and housing subsidiesalso exhibit a substantialimpact on
children.For all age groups,the measuredimpactof additionsto resourcesis larger
using datafrom the SIPPthandatafrom the CPS. As discussedearlier,this in part
reflectsbettercoverageof income sourcesand receiptin the SIPP survey.
Regardlessof the source of data,the overall impactof addingincome from the
varioussourceslisted to a measureof family resourcesis to lower povertyratesof
childrenmore than the rates of othergroups.Accordingto the CPS, the marginal
effect of includingall additionsin the familyresourcedefinitionis to lowerthe child
povertyrateby about20 percent,versusabout16 percentreductionsamongworkingage adultsand the elderly. With SIPP data,the decreasein the povertyrate is 38
percentamongchildren,versus about29 percentfor the othertwo groups.
Of the expenses listed in Table 2, subtractingmedical out-of-pocketcosts from
family incomes clearly has the largesteffect on povertyrates.Using CPS figures,
we find thatthe povertyraterises from 14.8 percentto 17.9 percentwhen medical-
7. Differencesin child-careexpensesbetweenthe two surveysresultfromdifferingmethodsof imputing
such costs. In the SIPP, we use actualreportsof child-careexpenses as indicatedin a wave six topical
moduleon child care,thoughthenonly subtractingexpensesfromthe familieswherebothparents(or the
parentin a single-parentfamily)workandwhich includechildrenyoungerthan12 yearsold. As the NAS
panelrecommended,we also put a cap on expensesequal to the maximumallowedunderchild caretax
credit,or the earningsof the parentwith the lower earnings(whicheveris lower). In the CPS poverty
measure,we deductchild-careexpensesby: (1) imputingwho incurredchild-carecosts amongfamilies
and (2) subtractinga
with workingparentsand childrenunder12 based on a few family characteristics;
weekly amount,basedon previousAFDC child-caredeductionguidelines,for each week workedby the
parentwho workedthe fewest weeks in the previousyear. Shortet al. (1999) containsmore details on
this method.Researchcontinueson variousmethodsof estimatingchild-careexpensesin experimental
povertymeasures(Iceland2000).
404
The Journalof HumanResources
Food
Child
School School
Energy
Mean
Mean
Mean
Mean
Mean
Housing
Mean MeanWomen,
care
stamps
Percentage
Percentage
Percentage
Percentage
Percentage
Percentage
Percentage
lunches
amount amount
amount amount amount
infants,
expense amount
assistance
subsidiesbreakfasts
with
expensesand
receivingreceivingreceivingreceivingreceivingreceiving
Table
Family
1
Benefits
and
expense
children
Expenses,
1992
assistance
NANA
CPS All
2,794
11.5
NANA 208
3.8
2,015
4.2
2,668
8.4
376
5.6
238
5.7
3,145
4.6
1,958
175
442
18.1
10.1
13.0
SIPP
2,431
8.5
NANA 210
2,159
2,460 NANA 612
50.1
48.3
18.5
17.6
CPS
1,890
4.0
2,554
578
233
3,806 240
418
64.7
56.7
28.1
37.2
25.6
22.6
SIPP
NANA 348
2,262
59.5
19.2
CPS
1,889
296
10.8
31.9
Persons
2,912 NANA 225
6.1
20.9
2,248
6.6
Poor
Persons
Children
241
2,745
400
8.7
13.8
10.2
3,541
7.4
195
2,384
500
22.1
35.3
20.3
NANA 660
NANA 223
2,505
2,561
2,514
65.5
72.9
13.1
23.0
21.4
1,960
5.4
2,931
431
626
4,057 256
238
78.5
79.0
52.3
30.7
32.3
30.4
SIPP
CPS
Poor
SIPP Children
Communications: Iceland, Short, Gamer, and Johnson
NA
Note:
Source:
benefits
Not
andmean
U.S.
405
State
Child Other
Social
MeanCapital
Mean Mean MeanFederal
Mean Mean
MeanEarned
MeanMedical
Unweighted
income
gains Percentage
Percentage
Percentage
Percentage
Percentage
Percentage
Percentage
Percentage
support
available.
N
income security
income
amount
Census
amounts
(oramount expense expense expense expense expense expense
expenses
with
with
with
with
paid with
tax with
taxes with
work-relate
arerefer
taxes
taxes out-of-pocket
Bureau,
to
losses)receiving
credit
the
expense expense expense expense expense expense
1993
gain/loss
measured
expenses
at mean
expenses
the
Current
amount
family
Population
among
level.
those
Survey
who
March
reported
supplement
having
and
such
1992
NANA
6,383 841
1,147
155,019
2,986 1,930 6,118 2,458
14.1
81.9
74.4
84.4
15.4
66.2
93.6
NANA 772
3,321 NANA 7,005 2,525 4,013
45,390
93.7
13.8
83.9
3.0
74.0
1,182
83.9
23,175
1,374
2.8
991
291
561
121
54.3
3.8
13.8
37.6
1,775 NANA 605
55.1
78.7
NANA
receipts/expenses.
5,242
Income
Also
and
note
NANA 146
598
886
2.1
38.5
57.3
1,782 2,750
1.4
76.4
Survey
of
that
Program
the
629
57.2
42,723
7,079 841
2,930 1,900 5,789 2,493 NANA 1,092
92.7
89.9
63.8
12.8
70.6
87.6
27.0
unit
of
NANA
12,983
Participation
3,299 NANA 6,559 2,443 3,908
774
90.3
3.2
92.9
73.6
25.4
1,173
90.3
analysis
here
is
1,548
9,596
2.5
611
983
424
140
13.3
58.9
2.2
51.6
1,751 NANA 616
59.1
77.3
people,
but
the
NANA 877
NANA 205
2,366
622
48.3
61.7
1.0
1,635 2,967
76.1
1.5
619
61.5
406
The Journalof HumanResources
All
The
Table
2
Effect
of
Poverty
Poverty
to
WIC
Food
School
School
Earned
Capital
Energy
credit
rates rate
Housing
poverty
additions
Percentage
additions
thresholds
stamps
definitions
gains subsidy
lunch
using using
rate
resources:
income
breakfast but
(or
the
subsidies
change
tax assistance
with
in
subsidy
official
all
adding official
losses)
subsidy
-17.6
NA 14.1
14.1 NA14.8
14.8
14.4 14.6
12.2
14.8
CPS
-32.5
NA
11.8
11.8
10.9
11.6
8.0 11.3 11.8
11.0
11.9
SIPP
Various
Elements
of
the
All
Persons
Experimental
Poverty
-19.9
NA
21.1 NA22.3
22.3
17.9
21.0
22.0 21.9
22.3
CPS
Measure
on
Children
-37.6
NA
11.8 17.8 18.8
17.3
17.7
18.9
18.9
18.4
19.0
SIPP
-15.8
NA
NA
11.9
11.9
10.0
11.4
11.4
11.6 11.8
11.9
CPS
-29.0
NA
6.7 8.9
9.39.38.79.39.28.6
9.4
SIPP
-16.3
10.8
12.9
12.9
NA
NA
12.8
12.4
11.4 12.9
12.9
CPS
Poverty
Rates,
1992
Adults
18-64
Elderly
-27.8
NA
6.3 8.7
8.68.57.08.78.78.1
8.7
SIPP
Communications: Iceland, Short, Gamer, and Johnson
407
NA
Source:
Not
Poverty
All
Poverty
State Child
Child
Other
Social
from Experimental with
Federal
Medical from
rates
available.
poverty
poverty
care
measure
Percentage
Percentage
Percentage
elements
Census
work
thresholds
income support
subtrations
rate subtractions
butofficial
security
official
with
income
poverty
expenses
Bureau,
resources:
change
change
paid
taxes
with
with
subtracting
expenses
rates
taxes
these
in
expenses
in
difference
taxes
rates
1993
all
new
resource
U.S.
Current
14.6
46.6
NA
21.7
15.9
14.9
14.9
17.9 15.7
15.2
11.9
58.3
NA
18.8
12.9 12.0
15.3
12.7
11.9
12.0
14.8 25.6
-2.7
21.7
36.9
NA 23.1
24.1
25.1
30.5
22.5
23.5
22.4
-2.2 18.5
-0.4
18.9
41.5
NA 22.1
26.8
19.3
20.7 19.0
20.2
19.0
31.9 15.7
-0.8
11.8
49.4
NA
14.1
17.8
12.9
12.9
12.2
12.0
12.0
-0.1
29.4 12.1
9.4
63.0
NA
15.3
10.3 9.511.7
9.5
9.410.2
1.1
62.8 21.0
13.0
73.1
NA
22.3
12.9
12.9
21.7 13.0
12.9
13.0
98.4 17.2
2.4
8.9
116.7
NA
18.8
8.7 8.718.4
8.78.78.7
Population28.4
-1.4
19.0
Survey
March
21.4 14.4
0.0
supplement
and
1992
Survey
of
Income
and
Program
Participation
408
The Journal of Human Resources
out-of-pocket costs are subtracted from resources. Also of note are the effects of
Social Security taxes and work-related expenses on poverty rates. They raise the
poverty rate from 14.8 percent to 15.9 percent and 15.7 percent, respectively, according to CPS figures. The pattern among children and working-age adults tends
to follow the overall pattern. Among the elderly, however, the only expenses that
have a substantial effect on poverty are medical out-of-pocket expenditures. The
impact of these is very large, as a relatively high proportionof the elderly use medical
services and incur such expenses (Short et al. 1999).
Subtracting expenses from family's resources tends to increase poverty among
children less than it increases poverty among other age groups, in part because of
relatively higher initial poverty rates among children, which produces smaller relative effects, and also the smaller impact of medical-out-of-pocket costs. According
to CPS data, subtracting expenses increases the child poverty rate by 37 percent;
the increase among working-age adults is 49 percent, and among the elderly it is
73 percent-the last figure again largely reflects the considerable effect of medical
expenses among the elderly.
Table 2 also shows poverty rates using the official income definition, but with
new thresholds, equivalence scales, and geographic adjustments to the thresholds,
as discussed in the methods section. The overall effect of the experimental thresholds
on poverty rates is modest for all age groups.
The last two rows in Table 2 display the result of the experimental poverty measure, by survey, when all the recommendations are implemented simultaneously.
The data from the CPS indicate that, for all age groups, poverty is higher under the
experimental measure, as defined here, than under the official measure. The CPS
experimental poverty rate is 19.0 percent-28.4 percent higher than the official
rate.
It is important that the percent change in the poverty rate using the experimental
versus the official measure is smaller for children than for working-age adults and
the elderly. Among children, the poverty rate rises from 22.3 percent under the official measure to 25.6 percent under the experimental measure, a 14.8 percent increase.
The corresponding increases for working age adults and the elderly are 31.9 percent
and 62.8 percent, respectively. Thus, this analysis confirms the notion that the restricted definition of resources under the current official poverty measure overstates
the material disadvantage of children relative to other age groups. This finding supports recent work by Betson and Warlick (1998) and Short et al. (1999), who find
that the poverty gap between children and the elderly is narrowing, not widening,
when an expanded definition of family resources is employed.
Yet it is important to note that child poverty rates under the experimental measure
still surpass the poverty rates of the other two age groups, even if the increase in
the experimental measure versus the official rate is smaller among children. The
CPS estimate indicates that a quarter of children lived in poverty in 1992.
Poverty rates for all groups are somewhat lower when measured with SIPP data.
Poverty rates under the SIPP experimental measure range from 12.1 percent for
adults to 18.5 percent among children, who once again display the highest poverty
rate. The more expansive definition of resources, coupled with thorough reporting
of noncash benefits in the SIPP, account for much of this.
Communications: Iceland, Short, Garner, and Johnson
Table 3
ChildPovertyRates by VariousCharacteristics,CPS 1992
All children
Race/ethnicity
Non-Hispanic white
Non-Hispanic black
Hispanic
Other
Family Type
Married-couple
Male-headed (unmarried)
Female-headed (unmarried)
Education of household head
Less than high school
High school
Some college
College graduate
Post-graduate
Parents' work status
No working parent present
Working parent present
Age of child
0-2
3-5
6-11
12-17
Official Poverty
Measure
Experimental
Measure
Percent
Difference in
Poverty rates
22.3
25.6
14.8
13.2
46.6
40.1
20.5
15.9
48.0
48.5
24.3
20.5
3.0
20.9
18.5
11.3
27.2
55.4
14.9
35.8
56.7
31.9
31.6
2.3
51.7
23.5
13.6
4.0
2.1
56.2
27.6
16.1
6.1
3.6
8.7
17.4
18.4
52.5
71.4
91.1
15.2
85.9
19.3
-5.7
27.0
27.3
25.5
21.6
18.6
32.6
30.1
24.1
20.8
19.4
18.0
11.6
11.8
Source:U.S. CensusBureau,1993 CurrentPopulationSurveyMarchsupplement
A. Child Povertyby SelectedDemographicCharacteristics
Table 3 displays poverty rates calculated for various subgroups of the population
for 1992 using the official and experimental measures estimated with CPS data.8It
shows that estimated experimental poverty rates are not as much higher relative to
the official measure among Black children than among others. Further analysis (not
shown) indicates that greater receipt of noncash government transfers and lower
average medical-out-of-pocket expenses of black families with children who are poor
under the official measure account for this.
8. Demographicpatternsare fairly similarwhen using SIPPdata.
409
410
The Journal of Human Resources
30.0
25.0
20.0g
15.0-
v
10.0 -
|._
Official poverty
|-
Experimentalpoverty
5.0
0.0
1992
1993
1994
1995
1996
Year
Figure 1
CPS Official and Experimental Child Poverty Rates, 1992-1996. Source: U.S.
Census Bureau, 1993-1997 Current Population Survey March supplements
The experimental measure also produces higher estimated poverty rates for children in married-couple families and single-parent male householder families, but
fairly similar rates for those in female-headed families with no spouse present, than
we see under the official measure. Married-couple and male-headed families have
considerably higher experimental poverty rates in part because they are less likely
to receive government transfers than single-parent female householder families.
Results also indicate higher estimated poverty rates under the experimental measure than under the official measure for children in families with workers and in
families headed by more highly-educated householders. These families tend to receive fewer government transfers and incur higher work-related expenses. They also
have relatively high out-of-pocket medical expenses.
B. Time Trends
We examined trends in poverty over the 1992 to 1996 period. Figure 1 shows that
the experimental and official poverty rates for children follow the same general trend,
rising in 1993 before falling for three consecutive years from 1994 to 1996. The
experimental rate fell at a slightly faster rate, however, mainly because the receipt
of benefits among children's families who are poor under the official measure rose
by more than the modest increase in nondiscretionary expenses over the time period.
By far the most significant increase in these benefits over the period occurred in the
EITC program. Poor children lived in families which received, on average, more
than $700 dollars more in EITC tax credits in 1996 than in 1992 (in constant 1992
dollars).
Communications:
Iceland,Short,Garner,and Johnson 411
IV. Conclusion
This analysisaddressesthe issue of how our view of child poverty
woulddifferif a new, refinedpovertymeasurewereadopted.We basethe implementationof the experimentalpovertymeasureon researchof the NAS Panelon Poverty
and FamilyAssistance(Citroand Michael 1995), and a subsequentCensusBureau
report(Shortet al. 1999).
We find that experimentalpovertyrates tend to be higherthan official poverty
rates,andthatchildpovertyrates,by any measure,continueto surpassthose of both
working-ageadultsandthe elderly.However,the gap betweenchild and adultpoverty rates is smallerunderthe experimentalmeasure.This reflectsthe findingthat
combinedeffect noncashgovernmenttransfersandthe EITChave a largerpovertyreducingeffect on childrenthanon others,andsubtractingexpensesproducessmaller
povertyincreasesamongchildrenthanothers.Of the governmenttransferprograms
not accountedfor in the official povertymeasurewe examined,food stamps,the
EITC,andhousingsubsidieshave the biggestimpacton reducingchildpovertyrates
underthe experimentalmeasure.The estimatedimpact of benefits is even larger
when using SIPP data than CPS data, in part due to more thoroughreportingof
benefitsin the SIPP.
We also findthatmovingto an experimentalpovertymeasurewouldhaveimplications for the compositionof the childpovertypopulation.Blackchildrenwouldcomprise a slightly smallerproportionof the child povertypopulation,in partbecause
of greaterreceiptof benefitsand lower medicalout-of-pocketexpenses than other
families.Similarly,the currentofficialpovertymeasuretendsto underestimate
poverty amongchildrenin married-couplefamilies,and overstatepovertyamongchildrenin female-headedfamilies,becauseof lowerreceiptof benefitsandhigherworkrelatedexpenses amongthe formergroup.
We also findthatthe experimentalandofficialpovertyratesfollow the samegeneral time trend,rising from 1992 to 1993 before falling from 1994 to 1996. The
experimentalratefell at a fasterratein the latterperiod,however,reflectingincreases
in the EITCprogram.In this way, the experimentalpovertymeasuremoreaccurately
capturesthe impactof both economic conditionsand governmentpolicy than the
currentofficial measure.
References
Betson, David. 1996. "Is EverythingRelative?The Role of EquivalenceScales in Poverty
Measurement."Universityof Notre Dame. Unpublishedmanuscript.
Betson, David M., and JenniferL. Warlick.1998. "AlternativeHistoricalTrendsin Poverty." American Economic Review 88(2):352-56.
Centeron Budgetand Policy Priorities.1998. "Strengthsof the Safety Net: How the
EITC,Social Security,and OtherGovernmentProgramsAffect Poverty."ResearchReport 98-020. Washington,D.C.: Centeron Budgetand Policy Priorities.
Citro,ConnieF., and RobertT. Michael,eds. 1995. MeasuringPoverty:A New Approach.
Washington,D.C.: NationalAcademyPress.
Coder,John,and Lydia Scoon-Rogers.1996. "Evaluatingthe Qualityof IncomeData Col-
412
The Journalof HumanResources
lected in the AnnualSupplementto the MarchCurrentPopulationSurveyand the Survey of Incomeand ProgramParticipation."WorkingPaper215, Surveyof Incomeand
ProgramParticipation.WashingtonD.C.: U.S. CensusBureau.
Dalaker,Joseph,and MaryNaifeh. 1998. "Povertyin the United States:1997." Report
P60-201, CurrentPopulationReportSeries. Washington,D.C.: U.S. CensusBureau.
Iceland,John.2000. "The Effect of Families'Child Careand Work-RelatedExpenseson
Poverty."Paperpresentedat the Joint StatisticalAssociationmeetings,Indianapolis,August 2000.
Ruggles, Patricia. 1990. Drawing the Line: Alternative Poverty Measures and Their Implications for Public Policy. Washington, D.C.: The Urban Institute Press.
Short,Kathleen,Thesia I. Garner,David Johnson,and PatriciaDoyle. 1999. "Experimental PovertyMeasures:1990 to 1997." ReportP60-205, CurrentPopulationReport,ConsumerIncome.Washington,D.C.: U.S. CensusBureau.