The Origins of African-American Family Structure

THE ORIGINS OF AFRICAN-AMERICANFAMILYSTRUCTURE*

STEVENRUGGLES Universityof Minnesota

I use a new data source, the Integrated Public Use Microdata Series, to trace race differences infamily structure between 1880 and 1980. Analysis confirms recentfindings that the high incidence among African-Americans of single parenthood and children residing without their parents is not a recent phenomenon. From 1880 through 1960, black children were two to three times more likely to reside without one or both parents than were white children. In recent years, however, the race differential in parental absence has grown. Also, blacks have had a consistently higher percentage of extended households than have whites, but until 1940 this was the result of single parenthood and parentlessness among children: Extended households were more common among whites once the effects of absent parents were controlled.

T wo distinctivefeaturesof African-Ameri- egories. Many theorists have argued that

can family structure are well known: Moynihangot it backwards:Thedisadvantaged

Blacks are far more likely than whites to be- position of blacks is not the consequence of

come single parentsand to reside in extended single-parentfamilies, but ratherthe cause of

families. The theoriesofferedto explain these them.This interpretationis frequentlyaccom-

differences are diverse and have generated panied by the thesis that the black extended

heated debates. I attempthere to narrowthe family has been a means of coping with both

field of potentialexplanationsfor race differ- povertyandsingleparenthood(Allen 1979;Bil-

ences in family structureby presentinga con- lingsley 1968; Farley 1971; Fischer,Beasley,

cise and detailed description of changes in andHarber1968;Hofferth1984;Rainwaterand

black family structurefrom 1880 to 1980.

Yancey 1967;Stack 1974). Othertheoristshave

Since the turnof the century,social theorists maintainedthat Moynihan's stress on pathol-

have arguedthatslaveryresultedin disorgani- ogy and disorganizationignores the resilience

zationandinstabilityin blackfamilies(DuBois of the blackfamily,andespecially the strength

1899, 1909; Elkins 1963; Frazier 1932, 1939; of extended kin ties among blacks (Aschen-

Myrdal 1944). This interpretationculminated brenner 1973; Hays and Mindel 1973; Hill

with Moynihan's (1965) report, which con- 1971;MartinandMartin1978;McAdoo 1980;

cluded that the "pathological"natureof black Riessman 1966; Staples 1975). These analysts

communitiescould be tracedto the deteriora- often point to persistent cultural differences

tion of black family life.

betweenwhites andblacks-originating in sla-

The numerousargumentsagainsttheMoyni- very or in African culture-to explain both

han reportcan be divided into two broadcat- single parenthoodand extended family struc-

ture (Nobles 1978; Owens 1976; Scanzoni

*Directcorrespondencteo StevenRuggles,De- 1971; Shimkin,Shimkin,andFrate1978). partmenotf History,267 19thAvenueSouth,Uni- The controversyover the Moynihan report versity of Minnesota,Minneapolis,MN 55455; stimulateda spate of revisionist historicalin-

internet: RUGGLES@ATLAS.SOCSCI.UMN.EDUF. unding vestigations into African-American family

fordatapreparatiownasprovidedby theNational structure. These studies asserted that black

Science Foundation(SES-9118299 and SES- families in the late nineteenth century were

9210903),the NationalInstituteof ChildHealth and HumanDevelopment(HD 25839), and the GraduateSchoolof the Universityof Minnesota. This researchwas carriedout undera BushSab-

overwhelmingly male-headed and nuclear in structure. Although some authors acknowledged minordifferencesin family structurebe-

baticalfellowshipfromthe Universityof Minne- tween blacks and whites, they all maintained

sota.My thanksto RobertMcCaa,StuartTolnay, thatin practicaltermsblack families were es-

andtheanonymouAs SRreviewersfortheirhelpful sentially similar to white families (Agresti

commentsandsuggestions.

1978; Bigham 1981; Carlson 1988; Fursten-

136

AmericanSociological Review, 1994, Vol. 59 (February:136-151)

THE ORIGINSOF AFRICAN-AMERICANFAMILYSTRUCTURE

137

berg, Hershberg, and Modell 1975; Gutman studies of African-Americanfamily structure

1975, 1976; Harris 1976; Krech 1982; Lam- have used national census data from the turn

mermeir 1973; Pleck 1972; Riley 1975; Shif- of the century,but even they have not yielded

flett 1975). The revisionists thus implied that statistics that are directly comparableto data

the distinctive African-Americanfamily pat- availablefor the recentpast.

ternis of recentorigin, andthis reinforcedthe Now, a new datasource,the IntegratedPub-

now widespreadview thateconomic disadvan- lic Use MicrodataSeries (IPUMS), allows us

tages faced by blacks in the recentpast arere- to generatefor the first time a consistent na-

sponsible (Brewer 1988; Wilson 1987).

tional series of statistics on family structure

A morerecentgroupof historicalstudieshas over the pastcentury(Ruggles et al. forthcom-

revised the revisionists'conclusion. These in- ing). The IPUMS is a nationalhistoricalcen-

vestigationshave stressedthe continuityof Af- sus databasein preparationat the Universityof

rican-Americanfamily structureby showing Minnesota under funding from the National

that the high frequencyof single-parentfami- Science Foundationandthe NationalInstitutes

lies among blacks observed by Moynihan in of Health.The IPUMScombines nationalcen-

1965 already existed around the turn of the sus microdatafrom a varietyof sources. Since

century (Brown and Kallgren 1989; Goeken 1960, the U.S. CensusBureauhas madepublic

1989; GordonandMcLanahan1991; Morgan, use microdatasamplesavailableto researchers

McDaniel, Miller, and Preston 1993; Ruggles within a few years of the decennial enumera-

andGoeken 1992). These studieshave madeit tion (U.S. Bureauof the Census 1972, 1973,

clear,therefore,thatthecurrentAfrican-Ameri- 1982). Since 1979, projectscarriedout at the

can family patternwas not a responseto recent University of Minnesota, the University of

economic changes. To explain the distinctive Washington,the University of Pennsylvania,

features of the African-Americanfamily, we and the University of Wisconsin have con-

need to look back at least as far as the nine- verted large national samples of the 1880,

teenth century.

1900, 1910, 1940, and 1950 population cen-

My investigationbuildson therecenthistori- suses into machine readable form (Graham

cal research. I describe for the first time the 1979; Ruggles et al. 1993; Stronget al. 1989;

long-termtrendsof single parenthoodandex- U.S. Bureau of the Census 1984a, 1984b).

tended family structure among blacks and Similarprojectsare now underwayat Minne-

whites, using a new seriesof census microdata sota for the 1850 and 1920 census years.Alto-

that spans the period from 1880 to 1980. The gether,large samples of census microdataare

recenthistoricalstudieshave shownthatsingle anticipatedfor atleast 11census yearsbetween

parenthoodand extended families were more 1850 and 1990.

common among blacks than among whites at These census files employ a wide varietyof

the turnof the century,but they have not been samplingstrategies,sample units, and coding

able to trace these changes over the course of schemes. The IPUMS imposes a consistent set

the twentiethcentury.

of definitionsandcodes on the data,establish-

ing order and maximizing the potential for

DATA

valid andreliableanalysisof long-termchange (Ruggles 1991, forthcoming b). When com-

Until recently,historianslacked adequatedata plete,the IPUMSwill includenationalsamples

to trace long-term national trends in family of coherentcensus microdatafrom all census

structure. With few exceptions, empirical years for which data are available. I used the

analyses of race differencesin historicalfam- preliminary version of the database in this

ily structurehave focused on a single moment analysis;-it includes census data from 1850,

in time. Moreover,the bulk of historicalstud- 1880, 1910, 1940, 1960, and 1980.1

ies have examinedliving arrangementsin one

or two communities.We have had no way of determiningif the communitiesarerepresentative, and comparisons between studies have

1 The IPUMS is scheduled to be released in the summer of 1995 through both the National Archives andthe Inter-UniversityConsortiumfor Political and Social Research. A preliminaryversion

been complicatedby variationsin datasources, of the data series is available from the author.For

data collection procedures,and classifications descriptons of the source data, see text citations.

of family structure.The most recent historical The sample densities used throughoutthis study

138

AMERICANSOCIOLOGICALREVIEW

Table 1. PercentageDistributionof HouseholdCompositionby Race: United States 1880-1980

FragmentaryHouseholds Primaryindividuals Single parents

Total fragmentary

Black 1880 1910 1940 1960 1980

8.5 10.5 15.4 17.6 26.7 11.7 9.7 8.7 9.9 20.5 20.3 20.2 24.1 27.5 47.2

White 1880 1910 1940 1960 1980

5.0 6.2 9.5 14.6 26.7 8.2 7.4 7.1 5.1 7.1 13.2 13.6 16.6 19.7 33.9

Married Couple Households

Childless couples

11.6 16.9 19.6 16.0 10.6

Couples with children

45.6 38.5 29.8 31.1 25.1

Total marriedcouple

57.3 55.4 49.4 47.1 35.6

11.0 14.5 20.5 23.2 24.7 56.4 51.9 45.6 45.6 34.9 67.3 66.5 66.0 68.8 59.6

Extended Households Vertically extended Other extended

Total extended

13.1 14.1 16.6 14.1 10.1 9.4 10.3 9.9 11.3 7.0 22.5 24.4 26.5 25.4 17.2

10.7 10.9 11.0 8.8 9.0 6.4

6.9 3.9 4.6 2.6

19.5 19.9 17.4 11.5 6.6

Total percent

Numberof households

Index of dissimilarity (Black-white, six categories)

100.0 100.0 100.0 100.0 100.0 12,449 8,616 6,276 4,789 8,387

10.7 13.4 16.7 21.7 24.0

100.0 100.0 100.0 100.0 100.0 84,398 70,375 62,678 47,879 69,024

Note: Excludes groupquartersunder 1970 Census definition. Primaryindividualsare persons heading households with no kin present;single parentsare unmarriedheads with childrenand no otherkin; childless couples are married couple households with no kin; couples with children are marriedcouples with children and no other kin; extended households arehouseholds with kin otherthanspouse andchildren;verticallyextendedarehouseholds with ancestors, descendants,or children-in-lawof head.

Despite the problemsof compatibilityintro- holder,maritalstatus,age, and sex-are virtu-

duced by theircompilers,the U.S. Public Use ally identical for all available census years

MicrodataSamples (PUMS) currentlyconsti- since 1880.2

tute the most consistent and comprehensive

source available for the study of long-term change in family and household composition. From 1850 onwards,the census definitionsof

LONG-TERMTRENDSIN HOUSEHOLD COMPOSITION

the basic units of enumeration-now called households-have varied only modestly, and the proportionof units affectedby such variations has been insignificant. The concepts of

Table1providesa generaldescriptionof household compositionby race from 1880 to 1980. The classificationscheme is a compromisebe-

quasi-household and group quarters, introducedsince 1940, have no clearanalogsin earlier censuses, but the PUMS files for all years providesufficientinformationto impose a low-

2 For discussions of the temporalcomparability of the census conceptsof family andhousehold,see Ruggles.(1991) and Smith (1992). The definition of groupquartersused here is the one used for the

est common denominatorof the groupquarters 1970 Census;thatdefinitionis the only on thatcan

concept.Moreover,the basicinquiriesrequired be appliedconsistentlyacrossall census yearsfrom

to classify household and family composi- 1880 to 1980. For the sake of consistency, persons tion-relationship to householdheador house- residingin groupquartersunder 1970 Census defi-

nitions were excluded from analysis in this study,

were 1/200 for 1850, 1/100 for 1880, 1/250 for except for Table 3 and for the statistics from the

1910, 1/500 for 1940, and 1/1000 for the remaining 1850 Census.On the potentialeffects of this exclu-

years.

sion, see Ruggles (1988, 1991).

THE ORIGINSOF AFRICAN-AMERICANFAMILYSTRUCTURE

139

tween the U.S. Census Bureau approach to markablystable, at least through 1960. This

household structureand the system developed findingsupportsthe recentstudiesarguingthat

by Laslett (1972) that is widely used by histo- the distinctive features of the African-Ameri-

rians.Householdsaredivided into threebroad can family have deep historicalroots.

categories on the basis of the composition of Despite this long-standing continuity, the

the primary family, which is defined as the race differentialin householdcompositionhas

group of kin related to the household head.3 notbeen static.InTable 1 the index of dissimi-

Fragmentaryhouseholdsconsist of primaryin- larity compares the distributions of black

dividualsandsingle parentsresidingwith their households and white households across the

children only. Marriedcouple households in- six detailed categories. From 1880 through

clude marriedcouples residing with no other 1980, divergence between black and white

kin and married couples residing with their household composition increased-the index

children and no other kin. Extended house- of dissimilarity rose gradually from 10.7 to

holds include additionalkin, such as parents, 24.0. Thus, althoughthe origins of the charac-

siblings, or grandchildrenof the household teristic patternsof black household composi-

head.Verticallyextendedhouseholdsarethose tion can be traced to the nineteenth century,

that include ancestors, descendants, or chil- race differences have become far more pro-

dren-in-lawof the householdhead.

nouncedover the course of the twentiethcen-

Among both blacks and whites, the most tury.

strikingchange in household structureshown

in Table 1 is the increasein the percentageof primaryindividualhouseholds-persons who reside alone or with nonrelativesonly. In the

THE LIVINGARRANGEMENTSOF CHILDREN

nineteenthcentury,blacks resided as primary Measurementby households minimizes race

individuals significantly more often than differencesin family structure.We can obtain

whites, but by 1980 this race difference had a clearer view by focusing on the living ar-

disappeared.The dramaticrise of the primary rangementsof children.There are several ad-

individualover the past centuryhas generated vantages to analyzing children living with

a largeliterature(see forexampleKobrin1976; single parentsinstead of single-parenthouse-

Ruggles 1988).

holds. To begin with, we can broaden the

Forthis analysis,the most importantcatego- analysis by looking at children who resided

ries in Table1 aresingle-parenthouseholdsand with no parentsas well as childrenwho resided

extended households.Among blacks, the per- with a single parent.Moreover,by measuring

centage of single-parenthouseholds was rela- the living arrangementsof children, we can

tively stablefrom 1880 through1960, andthen easily capture single parents who were not

it increased sharply. The percentage of ex- householdheads and who resided in subfami-

tended households among blacks was also lies and secondaryfamilies. Furthermorethis

fairly stable between 1880 and 1960, but has measurementstrategygreatlysimplifies analy-

droppedsignificantlysince then.

sis of measurementerrorresultingfromdemo-

In all census years, a smallerpercentageof graphicchange.4Finally, children are an im-

white households were fragmentary or ex- portantobject of study in theirown right, and

tended than were black households, and they arekey to understandingthe family as an agent

more often consistedof marriedcouples resid- of socialization.

ing with children. These race differences in- Table2 shows the percentageof childrenre-

creased between 1880 and 1980. The key cat- siding with both parents, with mothers only,

egories of black household structure, how- with fathers only, and without either parent.

ever,-single-parent and extended-were re- The overallpercentageof childrenwith one or

bothparentsabsentis summarizedin Figure 1. 3 These systems aredescribedin U.S. Bureau of

the Census (1983) andin Laslett (1972). Following

U.S. CensusBureaupractice,the termfamily refers 4For technical reasons, it is extremely difficult

here to any group of relatedpeople who reside to- to account for the effects of demographic factors

gether, whereas the term household refers to a on living arrangements when measurements are

group of people who shareliving quarters,regard- takenat the level of households (see Ruggles 1986,

less of their relationships.

1987).

140

o C.:

VI

_ _t

ZU

-t _ o

- oo 11.0N

't

r- x- o?

-_

7.,

:3

't'O r000 0

AMERICANSOCIOLOGICARLEVIEW

x-o ? N

C7,0

10

_

n

t-N

't

0

kn It N t-o"

:4 It

N _

00 N

m~~~~~~~~~~~O.0 O.O .

eOn . .

O .O

O en.O VI)

VI)

888o8

888

o8?8w8 i 1 8-88c ir- 88866 cr

N

0 Z

+ oO :: Xo m

.l .l _.l

__

1o.1mo l .

en

.~ Fio oo. , C. o _

WI V . ool x.l '_.l

_

m .l .

_

e Zo=

~~~OtneNtt 00

t

W) . o.o :

_

mm

tFo

nmo

__o

to

0

as~~~~~~~~~

3~~~~~o~710

00 00 _0 0_71 _CO10

_N 0

_ONC71

ON O_ N _0 00

o

00

c

00

r~~~~~~~~~ V)

O :: ;, e

N ?

o

~?~~-t~0_0~~0e0n~.

00

en 0 ?, _. 00 W) 0 OoN

00 xDb _ n_

Z

tn xoc r _ tn

xDm n F eF

o

o t X

:otof Fo

ox

U

Z

en t - en r-

m OoNob ON

ooo

0 Z

e ; R N ~~~o~~~~~~oo~0c o~~4 ~~~_~ci~_l _6 _ l _ _i

eOnoo

_

en

l

_11

o

_i

o N

m m

i

00

_

eW NoN?

i

mQ m O

t t-4 ^

n

Nt ' 00

V) O 00 00 m

.a:

3= ;^ _ _ O N ^

N w oNo o00.

mo0 .o

tn

t' N

NenwN 0 o

Xo~~~~~~~~~~~~~~I

cl

o,

ot

:~&~~~~~~~~~0~~~~~-

1. -4

r:

o 8

L

't to

_ o

do

DOr~~0 U

' o

7 o

"ot

o

0 4% _t

t

o

C)W) t 0

D

U

U

0

oo R

O

THE ORIGINSOF AFRICAN-AMERICANFAMILYSTRUCTURE

141

60-

L 50

5-C 0

0 C 400 0 1 30I co C

Blacks

Whites

C

am20-

10

10*

0 1880

1910

1940

1960

1980

Year

Figure 1. Percentagesof ChildrenAges 0 to 14 With One or Both ParentsAbsent, by Race: United States, 1880-1980

The presence of parentswas determinedfrom greaterover those two decades thanit was for informationavailable consistently from 1880 blacks:Thepercentageof white childrenresidto 1990: relationshipto head, age, sex, marital ing withoutbothparentsrose 85 percent,comstatus, and sequence within the households pared with only a 64 percent increase among From 1880 through1960, about30 percentof black children. Measuredas an absolute perblackchildrenages 0 to 14 residedwithoutone centage, however,the recent change has been or bothparents.By contrast,only about10per- far greateramongblacks, because they started cent of white children resided without one or from a higher base percentage. Parental abboth parentsover the same years. From 1960 sence among black childrenrose from 32 perto 1980, parentlessnessand single parenthood cent in 1960 to 53 percent in 1980, while rose sharply among both blacks and whites. amongwhite childrenthe increasewas from 9 The rate of change for whites was somewhat percentto 16 percentover the same period.

5 In the great majorityof cases, the determination of the parent-childlink was based entirely on family relationship and age. Occasionally, there was more thanone potentialmotheror fatherfor a given child; in such circumstances,the choice between alternateparentswas basedon maritalstatus, if possible,or proximitywithinthe householdif not. Detailed programminginstructionsfor the linking procedureappearin Ruggles et al. (1993). The percentage of childrenreportedin Table 2 as residing withoutmothersin the 1910 census year is slightly differentfrom the figures reportedin Morganet al. (1993). This is because the IPUMS uses a different procedure to link parents and children than that used in the 1910 project. The 1910 project used

some additionalvariableswhich more consistently identified step-childrenand adopted children (i.e., surname,detailed relationshipcategories, number of childrensurviving,numberof childrenever born, parentalbirthplaces,marriageage, and durationof marriage).Since this additionalinformationis not available across all census years, it could not be used to construct a consistent set of parent-child links. Accordingly,the linking procedureused was more likely thanthatused for the 1910 study to assign parenthoodto a few people who are actually step-parents or foster parents. Any such errors, however, should be consistent across all census years studied.

142

AMERICANSOCIOLOGICALREVIEW

Despite the dramaticchangesof recentyears, Table 3. Estimated Effects of Parental Mortality on

it is clearthatthe racedifferentialin childrens' residencewithparentsis notnew.In everycen-

Residence With Parents, by Race and Age: United States, 1880

sus year,the percentageof black childrenages

Blacks

Whites

0 to 14 living withoutone or bothparentswas Variable at least twice as high as the percentageamong

Fathers Mothers Fathers Mothers

white children.The percentageof childrenre- A. AssumedLife Expectancyat Birth

siding without either parenthas remainedre-

33.0 34.4 43.2 46.1

markablystable over the long run, at 10 to 12 B. EstimatedPercent of ChildenAges 0 to 14 Witha

percentforblacksand2 to 4 percentfor whites. Deceased Parent

Some of the historical differences between Ages Oto 4

4.6 4.0 2.8 2.1

blacks and whites in living arrangementsof childrenresultedfromdifferencesin mortality; children could not reside with parents who

Ages 5 to 9 Ages lOto 14

13.7 11.7 8.8 6.5 23.3 19.7 15.7 11.3

were dead. In the late twentieth century,few Total ages Oto 14 13.0 11.1 8.7 6.3

parents die before their children are grown. C. ObservedPercent of ChildrenAges 0 to 14 Withan However, parentalmortality in the ninteenth AbsentParenta

centurywas common.

Ages 0 to 4

23.2 13.9 6.6 4.5

Table 3 presents estimates of the effects of parentalmortalityon residencewith parentsin 1880. These estimatesshouldbe viewed as ap-

Ages 5 to 9 Ages 10 to 14

26.1 17.3 10.1 7.3 32.6 25.4 16.4 12.3

proximations,since ourknowledgeof racedif- Total ages Oto 14 26.9 18.3 10.7 7.8

ferences in mortalityin the late nineteenthcen- D. EstimatedPercent of ChildrenAges 0 Through14 tury is inexact. Panel A presentsthe assumed Witha LivingAbsentParent"

life expectancy at birth by sex for each race Ages 0 to 4

19.5 10.3 3.9 2.5

basedon life tablesfor the periodfrom 1870 to Ages 5 to 9

14.4

6.3

1.4

0.9

1880. Panel B shows the percentageof fathers

and the percentageof motherswho would be

Ages lOto 14

12.1

7.1

0.8

1.2

deadfor each age groupof childrenunderthese Total ages Oto 14 15.9 8.1 2.2 1.6

mortality conditions. Panel C shows the percentageof childrenwith absentfathersandabsent mothers.These figuresaresimilarto those in Table2, but they includechildrenwho were residentsof groupquarters.6Panel D of Table

a Includeschildrenresiding in groupquarters b This estimate was calculated as:

(Panel C) - (Panel B) 100 - Panel B

3 estimates the percentageof childrenwith a

survivingparentwho is absent,calculatedfrom Table3 furtherilluminatesthe findings pre-

panels B and C. The percentagesof children sented in Table 2. Among white children in

with a living absentparentmay be slightly un- 1880, parentalmortalitywas the main reason

derstated,because panel C includes some un- for the absence of parents:Only about 2 per-

knownpercentageof adoptedandstep-children cent of whites residedseparatelyfrom a living

residing with socially-defined parents whose mother or a living father.Despite the higher

biological parentswere dead.7

mortalityof blacks,on the otherhand,less than

half of the percentage of children residing

6 Figures in Panel C differ from the figures pre- withoutboth parentscan be accountedfor by

sented in Table 2 because Panel C includes chil-

drenresiding in groupquarters;unlike latercensus of theirchild until the time of the census. Parental

years, the 1880 sample allows full identificationof ages at the births of their children were tabulated family relationshipsfor persons in group quarters. directlyfromthe 1880 Public Use Sample for chil-

Since children residing in group quarterswere es- dren less than age 2 residing with a parentby the

pecially likely to have deceased parents,they must race and sex of the parent.The age distributionof be included for comparisonwith the parentalmor- childrenof each race was also tabulatedfrom the

tality estimates.

1880 sample. The life tables for blacks were taken

7 I used a simple life table approachto measure fromEblen(1974). Forwhites, I used a modelWest

parental survival, which involved calculating the level 11.45 life table (Coale and Demeny 1983), a

probabilityof death for each parentfrom the birth level recommended by Michael Haines for adult

THE ORIGINSOF AFRICAN-AMERICANFAMILYSTRUCTURE

143

4-4 0 s..,

z u

0

F-4 C,

0

F. 0

0

Z

'o

a +..- 7

1.4

0

0

Z

u

CU P"N

5 0

O

Cl)

C) Cl) 7

0

>-N

4-4

7

0

7 u

u z

0

7

C,

cl

0

F. 0

0

tell 4.4

64 7

z

64

+..0

4-4

0

0

Z

U

cl

tD

to

tw

>

0 C)

0 0

0

cl 0 tD

CT

tn

ON

'IC

tn

CT

ON

-4

00

en

CN

crr-

O O0 O0 O0

ON

tn

'IC

C

O00

ON

4

en CT CN en

0

4--,

O

tn 00

ON

+0.,

116 06 116 7

ON 00 r- 00

CeTn

00

tn

CT

'IC

CtnT

en CT CT 00

CIS

cl

.0 +C4

0 04 7

7

c; c;

00

en

tn

0

u =1 0

N

en

CT

O

CT

cl

7

w

tD

N

00

'IC

CT

CN ON

cl 1 6 cli

00

40. -4 0

7

0

rA 0

bcDl

0 0 64

0

ON

0 0 0

+-, O

+-,

tn

O 1-4

0

Cl)

C)

parentaldeath. Once we controlledfor the effects of mortality,parentalabsence was over five times more frequentamongblacks thanit was among whites.

Theracedifferencein residencewithparents in the late nineteenthand early twentiethcenturieswas mostpronouncedamongvery young children. Overall, for example, fewer than 5 percentof U.S. childrenunderage five had a deceased fatherin the late nineteenthcentury; but among blacks, 23 percent under age five residedwithouta father,comparedwith 7 percent of whites. The large race differential among the very young suggests that the African-American pattern of residence with a single parent did not usually result from the departureor death of a parent;in most cases, only one parentwas presentfrom a very early age. Indeed, Panel D of Table 3 suggests that the percentageof absentparentsamong black children with surviving parents actually declined with age, althoughthis finding may be an artifactof exaggeratedblack mortality.8

Wecanpushthe analysisback30 yearsprior to 1880 by comparingthe living arrangements of freeblackchildrento white childrenin a nationally representativesubsampleof the 1850 PUMS currentlyin preparationat the University of Minnesota(Menardet al. forthcoming). Unlikethe morerecentcensuses,the 1850 enumerationdid not provide explicit information aboutfamilyrelationships.I have thereforedeveloped a system of rules for inferringfamily relationshipson the basis of age, sex, surname, and sequence in the household. I tested these

mortalityin the period 1870 to 1880 (personalcommunication, 21 Apr. 1993; cf. Haines 1979). One might assume thatthe effects of parentalmortality could be assessed by looking at widowhood among single parents, but as Preston, Lim, and Morgan (1992) have demonstrated,the maritalstatus variableis too unreliablein historicalcensus datato use it for this purpose.(On the problemof distinguishing biological parents from socially-defined parents, see note 5).

8 Using data on child survival from the 1900 Census, Preston and Haines (1991) recently demonstrated that demographers have overestimated black child mortalityin the late nineteenthcentury. If we assume that black adult mortality has been similarlyoverstated,the decline of parentalabsence with increasing age of childrendisappears.Lower black mortality would also imply a higher frequency of parentalabsence among black children with survivingparents.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download