Structural Determinants of Men's and Women's Personal Networks

Author(s): Gwen Moore

Source: American Sociological Review, Vol. 55, No. 5 (Oct., 1990), pp. 726-735
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2095868
Accessed: 01/11/2010 05:46

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.

http://www.jstor.org
 STRUCTURAL DETERMINANTS OF
MEN'S AND WOMEN'S PERSONAL NETWORKS*

GWEN MOORE
State Universityof New Yorkat Albany

Men's and women's personal networks often differ in composition, with women's more
focused on family and men's on nonkin, especially coworkers. Using data from the 1985
General Social Survey, I find that these gender differences arise in part from dissimilar
social structural locations of men and women, which lead to distinct opportunities for
and constraints on the formation of close personal ties. Most gender differences in
network composition disappear or are considerably reduced when variables related to
employment,family, and age are controlled. However, some gender differences remain.
Women have a larger number, higher proportion, and greater diversity of kin ties in their
personal networks than men, even when compared with men in similar social structural
positions.

Over thepasttwo decadesextensiveresearch suggest that women are less able to use net-
on social networks has demonstratedthe
importance of networks in diverse facets of
social life, including social support (e.g.,
Kadushin 1982), employment (Granovetter
1974, 1982; Lin, Ensel and Vaughn 1981; Lin
1982), and power and influence in organiza-
tions, communities,and nations (Laumannand
Pappi 1976; Moore 1979; LaumannandKnoke
1987; Miller 1986). Indeed, networkties have
frequentlybeen described as social resources
thatoffer valuablesupport,acquaintances,and
information(Lin 1982; McPhersonand Smith-
Lovin 1982, p.884, 1986; Campbell,Marsden
and Hurlbert1986).
Studies of personal networks, comprised of
an individualand the othersto whom he or she
is connected,have found that women and men
usually have networksof similar size (Fischer
1982; Marsden 1987). These studies have also
found, however, large gender differences in
networkcomposition.When comparedto men,
women have fewer ties to nonkinandmore ties
to kin, while men include more coworkers in
theirnetworks(Fischerand Oliker 1983; Well-
man 1985; Marsden 1987). These findings
*Directcorrespondence
to GwenMoore,Depart-
mentof Sociology,Social Science340, SUNY at
Albany,Albany,New York12222.Thispaperwas
presentedat the 1990meetingof theAmericanSo-
ciologicalAssociation.I thankRichardAlba,Wal-
ter Ensel,CharlesKadushin,GinaWan-foonLai,
GlennaSpitze,andASR editorsandreviewersfor
helpfulcommentson previousdrafts,GinaWan-
works as instrumentalresources and that men
benefit more from the diverse and extensive
networksuseful in findingjobs and advancing
their careers. Much research on networks as
instrumentalresourceshas focusedon weakties.
But organizationalstudies, usually measuring
strongerties, have confirmedfindings of gen-
der differences, often finding that men have
more extensive ties than women, especially to
powerfulpersonsin workorganizations(Miller
1986; Brass 1988).
Observed gender differences in networks
have sometimes been attributedto contrasting
dispositions of men and women toward inter-
personalrelationships,concludingthat women
are more disposed to maintainingcloser ties to
kin andfewerties outsidethefamily(e.g., Miller
1976; Chodorow 1978; Gilligan 1982). In con-
trast, the structural perspective has ascribed
gender differences in networks to the dissimi-
lar social structurallocations of women and
men (Fischerand Oliker 1983).
While researchersfrequentlyhave adopteda
structuralperspective to explain network dif-
ferences, the structuralperspective theory has
yet to be subjectedto a comprehensiveempiri-
foon Lai and Jeff Torlina for research assistance,
and Sharon Baumgardner,Joan Cipperly, and Ei-
leen Pellegrinofor manuscriptpreparation.The data
from the 1985 NORC General Social Survey were
obtainedwith the assistanceof BrianFisherfromthe
Center for Social and Demographic Analysis at
SUNY Albany.

AmericanSociological Review, 1990,Vol. 55 (October:726-735) 726

DETERMINANTSOF MEN'S AND WOMEN'S NETWORKS
cal test. Previousstudiesof differencesbetween
men's andwomen's networkshave largelyused
local samples and a few structuralvariables
(Fischer and Oliker 1983; Campbell 1988;
Wellman 1985; Gerstel 1988). In this study, I
investigated structuraldeterminantsof men's
and women's networks to identify structural
factors and to determineif gender differences
in networkscan be explained by these factors.
RESEARCHHYPOTHESES
Blau's axiom, that "social associationsdepend
on opportunitiesfor social contact" (1977, p.
281), summarizes the structuralperspective.
Social relations occur within an opportunity
context that precludes or makes possible vari-
ous kinds of social contacts (Blau 1977,p.79;
Marsdenforthcoming,p. 1). FischerandOliker,
reportingon data from a study of friendship
ties in northernCaliforniacommunities,argued
thatstructuralopportunitiesand constraintsare
the primarycause of genderdifferencesin net-
works:"Weproposethatthe differingpositions
of women andmen in the workforce, in marital
roles, and in parenthoodcreatedifferentsets of
opportunitiesfor andconstraintson friendship-
building"(1983, p. 130; also see Fischer 1982,
pp. 253-255; Wellman 1985; Gerstel 1988).
Structuralopportunities,particularlyjobs out-
side of the home andhigherincome, occurmore
often for men than for women. On the other
hand,structuralconstraintson the development
of networkties, suchas responsibilityfor house-
work and childcare, are experienced more of-
ten by women.
Importantstructuralvariables are those re-
lating to work, family, and age. Work-related
factors,includingpaidemployment,educational
attainment, and income, are positively (and
often strongly)relatedto networksize and ties
to nonkin;they increasethe opportunityto form
ties with personsoutside the family and neigh-
borhood(Fischer 1982, pp. 251-252; Marsden
1987, p. 129).
Although key factors in the structuralper-
spective relate to paid employment, Fischer
(1982) did not find occupationto be relatedto
network ties. Nevertheless, it is probablethat
some occupations offer more opportunityand
incentive thanothersto form ties with cowork-
ers. Tightly knit occupations such as typogra-
phers and coal miners have been found to fos-
ter comradeship among coworkers (Fischer
1982, pp. 104-05; Lipset, Trow and Coleman
727
1956). Consideringoccupationalgroupsrather
thanspecific occupations,high-statuspositions
may lead their occupantsto form more ties to
coworkers than do low-status positions. This
reflects the value of ties among occupants of
high-statuspositions for careermobility, work
conditions permittinginteraction,and the lim-
ited time these persons have for purely social
relationships.
Family structureand age also affect network
composition.The presenceof childrenat home
constrainsthe formationof ties, especiallythose
to nonkin who are not neighbors.Persons who
aremarriedhave moreties to kin andneighbors
and fewer ties to othersthando unmarriedper-
sons (Fischer 1982, p. 253; Wellman 1985;
Gerstel 1988; Hurlbertand Acock 1990). Net-
works also differ by age: Nonkin ties peak in
the earlythirtiesandgenerallydecline afterthat
(Fischer 1982, p. 253; Fischerand Oliker 1983;
Marsden1987, pp. 128-129).
Yet, in one respect,the structuralperspective
allows a morefundamentalrole for gender.The
work and family variablesthat shape network
structuresfrequently affect men and women
differently.As a result,genderand social struc-
tural variables can be expected to interact in
theireffects on personalnetworks(Fischerand
Oliker 1983). Women are not only less likely
to be employed outside the home and to have
lower individual incomes than men, but they
also generally retainresponsibilityfor the ma-
jority of housework and childcare even when
they are employed full-time; men's contribu-
tion to housework is relatively unaffected by
employmentstatus(Epstein 1988, pp. 209-212;
Berk 1988). Thus marriage and parenthood
often constrainwomen's opportunitiesto form
network ties to nonkin outside the neighbor-
hood, while marriageoffers men the time and
opportunityto form networkties beyond local
andkin boundaries,andhavingchildrenproba-
bly affects men's networks little (Fischer and
Oliker 1983, p. 129; Wellman 1985; Gerstel
1988; Campbell 1988, pp. 191-194).1An addi-
I Ties with kin are readymade,an advantage to
persons with scarce time to develop new relation-
ships (Wellman 1985, pp. 174-176). Kin may also
assist with childcare,and thus among families with
children at home may be in more frequentcontact.
In addition,parentsare likely to have a largernum-
ber of ties to neighbors' as a result of both neigh-
bors' proximity and the propensity of children to
form local friendshipsthat bring their parents,par-
ticularlytheirmothers,in contact with neighbors.
728
tional effect of women's household responsi-
bilities may be thatfull-time paid employment
leads to a decline in ties of all types other than
to coworkers,especiallyto nonkin.Among men
paid employmentmay increase ties to nonkin,
especially coworkers,while having little effect
on kin ties.
Although network structureshave various
components, the focus in this study is on net-
workrangeratherthanotherstructuralfeatures,
such as density. A broad range of ties, with
many strong or weak connections to diverse
others, is often seen as a valuableinstrumental
resource,while networkdensityis moreclosely
associated with social support(e.g., Campbell,
Marsden and Hurlbert 1986; Marsden 1987;
Campbell1988, p. 181;Kadushin1982). Range
has been conceptualized in various ways, in-
cluding volume of contacts and diversity of
alters (Burt 1983; Campbell, Marsden and
Hurlbert1986;Campbell1988).I measuredvol-
ume of contacts as the total numberof persons
namedas well as the numberof personsof each
type (e.g., kin or coworkers)in the network.A
networkwould have a greatervolume of ties if
it had a largeroverall size and a largernumber
of specific types of ties. I measuredthe diver-
sity of alters in two ways: (1) the relative pro-
portionof kin in the network,and (2) the num-
ber of different nonkin and kin types in the
network.Networkswith morediversityof types
of ties have greaterrange and indicate integra- engage in shared activities (Caldwell and Peplau
tion into diversesocial spheres.(Marsden1987). 1982; Aukett, Ritchie and Mill 1988). This argu-
DATA AND METHODS
The 1985 General Social Survey, a national
probability sample of 1534 English-speaking
Americanseighteen years old and older, gath-
ered data on respondents' backgrounds,atti-
tudes, and other variables,includingmeasures
of personalnetworks(Davis and Smith 1988).
Respondentswere asked to name people with
whom they had discussed "importantmatters"
over the past six months.2 The interviewers
2 The wording was, "From time to time, most
people discuss importantmatterswith otherpeople.
Looking back over the last six months - who are
thepeople with whom you discussed mattersimpor-
tantto you? Justtell me theirfirst names or initials."
If fewer thanfive nameswere mentioned,interview-
ers were instructedto probe,with "Anyoneelse?" A
sex bias in the wordingof the networkquestionwas
suggestedby anASRreviewerdue to women'sgreater
propensityto talk with theirassociates and men's to
AMERICANSOCIOLOGICALREVIEW
specified neitherthe contentof the discussions
nor the number of persons to be named
(Marsden1987, p. 123;Burt 1985, p. 119). The
majority of the 1531 respondents with valid
networkdata (85.6%) named between one and
five persons,8.9 percentnamedno one and 5.5
percentnamedsix or morepersons.For each of
the first five persons named, additionalques-
tions focused on their relationshipsto the re-
spondentand otherinformationas well.
Because persons named were likely to be
those to whom respondents felt close, these
networkdataarebest seen as measuresof strong
ties. For most respondentsthese were subsets
of a more extensive networkof strongties - a
subsetrepresentingthosepersonswho firstcame
to mind when thinkingaboutrecentdiscussions
of importantmatters.3For the small proportion
of respondentswho namedmore thanfive per-
sons, the data on reportedties are incomplete,
omittingrelationshipsto the sixth and, for sev-
enteen respondents,personsnamed later.
Respondentswere askedto indicatethe ways
they were connected to each of the first five
persons they named, specifically: spouse, par-
ent, sibling, child, other family, coworker,
member of group to which they belong (such
as churchorclub),neighbor,friend,professional
advisor or consultant, other nonkin. The first
five relationships were coded as kin and the
last six as nonkin.4In orderto clearly separate
ment for the inflation of women's network size,
however, is not supportedby the resultsof network
research using differently worded questions. Two
previousstudiesgatheringdataon ties involvingboth
discussions and mutualactivities found, as does this
one, that women and men have networksof similar
mean size (Fischer and Oliker 1983; Gerstel 1988).
The more inclusive name generatorsused in those
studieswith less focus on close ties, however,yielded
far largernetworks,with a mean of 18 for both gen-
ders in the two studies.
I In a study of the personal networks of 33 per-
sons in Torontothatgathereddataon all individuals
with whomrespondentswere"significantlyin touch,"
Wellman found a median of four intimate ties as
well as seven less intimateones (1985, pp. 164-165).
His results suggest that most persons do not have
more thanthe maximumof five close ties described
in the NORC data.
4The wording was: "Hereis a list of some of the
ways in which people are connected to each other.
Some people can be connected to you in more than
one way. For example, a man could be your brother
and he may belong to your churchand be your law-
DETERMINANTSOF MEN'S AND WOMEN'S NETWORKS 729

kin and nonkin and to distinguish the more Table 1. Network Size and Composition Differences
voluntary nonkin ties, persons who were de- Between Men and Women
scribed by the respondent as being connected
Men Women
both by kinship and other relationships were
coded only as kin. Nonkin were all network Variables Mean S.D. Mean S.D.
memberswho were not describedas kin. Overall networksize 3.00 1.83 3.02 1.73
I used 12 dependentvariablesin this analy-
Absolute compositionb
sis. One, networksize, was the total numberof
# of kin 1.50 1.27 1.81** 1.34
persons named.5Eight variables reflected the
# of nonkinc 1.70 1.49 1.40** 1.32
"absolute"composition of each personal net-
work. These were the numbersof kin, nonkin, # of neighbors .19 .58 .26* .60
neighbors, friends, group members, advisors, # of friends 1.41 1.40 1.27* 1.28
coworkers,and othernonkin.6Three measures # of coworkers .62 1.02 .36** .75
of "relative"composition were included: the # of groupmembers .32 .79 .29 .75
proportionof kin, the numberof differentkin # of advisors .31 .73 .18** .52
types, andthe numberof differentnonkintypes # of others .06 .33 .04 .23
in the network. Relative composition
Obviously, the key independentvariable in Proportionkin .51 .38 .58** .36
this analysis was gender (male = 1). The other # of kin types 1.17 .88 1.43** .95
independentvariables can be categorized into # of nonkintypes 1.56 1.31 1.41* 1.28
one of four groups: (1) Variables concerned
withfamily structureincludednumberof chil- *p < .05, two-tailedt-test p < .01, two-tailedt-test
dren under 19 years old in the household, and aNetwork size is computed for all respondentsexcept
three for whom these data were missing (N = 1531). For
marital status (currentlymarried= 1). (2) Age persons with networks of size 6 or more, the number is
variablesincludedtherespondent'sage in years, recodedto the mean of 6.5.
andalso age squared(age2) to test for suspected bThe absolute and relative composition measures are
nonlineareffects. (3) The numberof years of based on all persons with networks of size one or larger.
education was the only variable in the educa- Thosenaming no one areexcluded(N = 1395).Therangefor
absolutemeasuresis 0 to 5.
tion category.(4) Finally,employmentvariables c Respondentscould reportmultiplerelationswith each
included employmentstatus (a set of dummy networkmember.For example, one tie could be described
variablesrepresentingfull time, parttime, and as a brotherwho is also a neighbor. In the calculation of
no employment),personal income (in 17 cate- number of nonkin of each type, persons who are also
gories, each recoded to equal the midpoint in describedas kin are excluded.
thousands),and type of occupation (a dummy
variablewhereprofessional/managerial = 1 and RESULTS
all other occupations = 0). Unfortunately,the
datadid not providemeasuresof the amountof Table 1 presentsthe means and standarddevia-
time devoted to householdlaborand childcare. tions for men and women on the variables
measuringnetwork size and absolute and rela-
yer. When I read you a name, please tell me all the tive networkcomposition.7While overall men
ways thatpersonis connectedto you."This question and women cited the same numberof persons
was followed by the probe:"Whatother ways?"
I After recoding,networksize had a range of 0 to in theirnetworks,they differedin the predicted
6.5. NORC coded all responses of 6 or more ties as ways on most othernetworkmeasures.In terms
6. This included a total of 84 persons, of whom just of absolutenumbersof personsof varyingrela-
17 named seven or more ties. Following the practice tionshipsto respondents,men's networks,when
of Marsden (1987, p. 126), I recoded the category
I The issue of selectivity bias has been raised as a
including all networks of size 6 and larger to its
meanvalue of 6.5. Because few respondentsreported potential problem in the analysis of network data
networks of more than five persons, this truncation when certain cases are systematically excluded on
was unlikely to seriously bias the analyses. the basis of the dependent variable (Marsden and
6 While the absolute composition measures in- Hurlbert 1987). In this analysis persons with net-
cluded counts of the number of nonkin of various works of size 0 were excluded from analyses other
types (e.g., friends and coworkers),it was not logi- than those of networksize. However, because these
cal to include such measuresfor kin since the num- persons made up less than 10 per cent of the total
berof specific types of kin (e.g., parentsandspouses) sample, selection bias was not likely to be a serious
is limited. problemhere (Marsdenand Hurlbert1987, p. 345).
730 AMERICANSOCIOLOGICALREVIEW
Table 2. Regressionof Network Size and Compositionon Genderand StructuralVariables

Absolute Composition Relative Composition

Category/ # #
Independent Network # # # Co- # # # % Kin Nonkin
Variable Size Kin Nonkin Ngbrs. workers Friends Group Advisors Kin Types Types

Gender -.248** -.278** .156* -.014 .016 .049 .030 .141** -.042* -.279** .062
(male = 1) (-2.62) (-3.59) (1.97) (-.39) (.32) (.63) (.65) (3.74) (-1.99) (-5.32) (.83)
Age
Age (years) -.015 -.043 .040** .005 .017* .025* .004 .015* -.013** -.035** .010
(-.96) (-3.40)** (3.06) (.92) (2.07) (1.97) (.56) (2.49) (-3.73) (-4.01) (.79)
Age2 a -.039 .397** -.481** -.049 -.202* -.354** -.026 -.149* .145** .290** -.176
(-.25) (3.00) (-3.55) (-.79) (-2.37) (-2.68) (-.33) (-2.32) (3.97) (3.23) (-1.38)
Education .134** .021 .065** .017** -.007 .048** .041 ** .026** -.011** .011 .059**
(years) (7.81) (1.50) (4.44) (2.62) (-.73) (3.39) (4.86) (3.75) (-2.75) (1.17) (4.32)
Employment
Full-time -.158 -.190 .019 -.093 .305** -.010 .016 -.061 -.039 -.116 .094
employment (-1.21) (-1.78) (.17) (-1.88) (4.44) (-.10) (.26) (-1.19) (-1.33) (-1.61) (.92)
Part-time .209 .072 .165 .007 .267** .166 .290** .044 -.014 .034 .453**
employment (1.36) (.57) (1.28) (.12) (3.31) (1.33) (3.90) (.72) (-.39) (.40) (3.76)
Occupation .207 -.034 .299** -.017 .238** .200* .066 .111* -.043 .031 .334**
(1.85) (-.38) (3.22) (-.41) (4.07) (2.21) (1.23) (2.52) (-1.72) (.51) (3.83)
Incomed .011* -.004 .010** -.003 .012** .008* -.002 -.000 -.002 -.001 .004
(2.41) (-1.12) (2.83) (-1.75) (5.37) (2.34) (-.84) (-.25) (-1.66) (-.49) (1.29)
Family
Maritalstatus .161 .646** -.649** -.045 -.023 -.580** -.041 -.023 .206** .690** -.318**
(married= 1) (1.75) (8.53) (-8.36) (-1.27) (-.48) (-7.68) (-.92) (-.64) (9.85) (13.45) (-4.35)
Children -.093* -.063 .020 .049** -.007 .019 .041* .017 -.008 -.014 .062
(#under 19) (-2.32) (-1.90) (.58) (3.17) (-.33) (.57) (2.09) (1.03) (-.82) (-.62) (1.93)

Intercept 2.091** 2.388** .092 -.040 -.131 .545 -.405* -.492** .902** 1.849** .562
(5.25) (7.38) (.28) (-.26) (-.63) (1.69) (-2.11) (-3.13) (10.11) (8.43) (1.80)
R2 .154 .078 .159 .027 .173 .120 .048 .050 .126 .154 .127

*p < .05 *p < .01

a
Coefficient multipliedby 1000.
b Employmentstatusis a set of dummy variables:full-time = 1, other = 0; part-time= 1, other = 0; not employed full-
time or part-timeis the omittedcategory.
c Managerialand professionaloccupations= 1, other occupations= 0.
d Personalincome was originallycoded in 17 categories.The midpointof each, in thousands,is used.
Notes: Unstandardizedcoefficients, t-statisticsare in parentheses.

comparedto women's, consisted of fewer kin lute networkcomposition,andrelativenetwork

and more nonkin, and included fewer neigh-
borsbutmorecoworkers,advisors,andfriends.
In relative composition, women's networks,
when comparedto men's, incorporateda larger
proportionof kin overallas well as morediffer-
ent types of kin, but fewer different types of
nonkin. All but two of the mean differences in
Table 1 are statistically significant at the .05
level or greater.
The remainderof my analysis focused on
structuraland gender determinantsof network
composition. Measuresof network size, abso-
composition were regressed on gender and
social structuralvariables.The resultsareshown
in Table2.8 Becausepastresearchhas suggested
thatsocial structuralvariablesandgenderinter-
act in the formationof networks, interactions
were systematically tested in all equations. A
global test was employed in each case by intro-
ducinga completeset of producttermsbetween
8 The "othernonkin"measurewas not examined
further, since, as a residual category, its meaning
was unclear.
DETERMINANTSOF MEN'S AND WOMEN'S NETWORKS
gender and all other independent variables.
Addinginteractiontermssignificantlyimproved
R2 for three of the 11 dependent variables:
numberof kin, numberof kin types, and num-
ber of advisors.Models including all of the in-
teraction terms are presented in Table 3. Be-
cause of collinearity, interpretingthe individ-
ual coefficients for the interactionterms was
somewhatdifficult. Therefore,I tested models
in whichI includedinteractionswith each group
of variablesseparatelyandthencombinedthose
groups that produced significant increases in
R2.These are presentedas the "trimmed"mod-
els shown in Table 3. Note that althoughnone
of the interactionswith the individualemploy-
ment variableshad a significanteffect on num-
ber of kin, for example, taken together they
significantlyimprovedthe R2of the model as a
whole.
NetworkSize and AbsoluteNetwork
Composition
On the whole, gender differences in network
composition were considerablyreduced when
structuralvariableswere controlled.Abouthalf
of the differencesthatwere statisticallysignifi-
cant in Table 1 were not significantin Table 2.
One of the remainingdifferences, the naming
of nonkin as networkmembers,had its magni-
tude reduced by half. The main exceptions to
this patternwere network size, where gender
differences were significant only in Table 2,
and the number of kin. Women apparently
maintainclose ties to a largernumberof per-
sons, especially kin, than do similarly situated
men.9On the other hand, men in general have
more ties to advisors and to nonkin.
Equally telling is the limited numberof in-
teractioneffects involving gender. Significant
interactionsoccurredonly in the equationsfor
kin, kin types, andadvisors.Suchresultslargely
refute the expectationthat similar social struc-
tural positions have different impacts on
women's and men's networks. For example,
the predictionthat marriagewould, in general,
increase men's nonkin ties while decreasing
women's was not supported(with the modest
I This statementis not affected by the reversalof
gender's sign between the kin equations with and
withoutinteractions.The positive gendercoefficient
in the equationwith interactions(Table 3) does not
indicate that men named more kin; it only partly
counterbalancesthe negative impact of the age in-
teractionsthat apply to men but not women.
731
exception of the numberof advisors). Nor did
the presence of children have a differential
impact on the absolute composition of men's
andwomen's networks.Of the interactionsthat
were significant, the most importanthad to do
with the numberof kin. One significant inter-
action was between genderand the categoryof
employment-relatedvariables (F4,1331= 2.763,
p < .05). More refined analysis (not presented)
suggests that this difference had to do with
employment status: Full-time employment
among women reducedthe numberof kin ties
named;no similareffect was foundamongmen.
Another significant interaction involved age
(F2,1331
= 4.698, p < .01) which affected the
number of kin named among men, but not
among women. The interactionof gender and
family variablesin the advisors equationindi-
cated thatmarriedmen and those with children
(but not women with the same family statuses)
namedmore advisorsamong their close ties.
Social structuralvariablesclearly had more
importantoverall effects than gender. Earlier
research found that family, employment, and
age variables have zero-orderrelationshipsto
networkstructures(e.g., Fischer 1982; Fischer
andOliker 1983; Marsden1987;Gerstel 1988).
Each of these structuralvariables played an
importantrole in some of the equations. For
instance, age had ratherconsistent and nonlin-
ear effects. Nonkin ties tended to rise during
the young-adultyears and then fall after that
point, while kin ties (only among men) dis-
played the opposite pattern.Maritalstatushad
especially large effects. Compared with cur-
rently unmarriedpersons, marriedpersons in-
cludedmorekin andfewer nonkin(particularly
friends) in their networks (see Hurlbert and
Acock 1990 for similarfindingsusing the same
data set). The expectation that children in the
householdwould have an effect similarto mar-
riage - increasing ties to kin and neighbors
while decreasing those to nonkin who are not
neighbors- was not confirmed.Childrendid
increase their parents' ties to neighbors (and
groupmembers),buttheirpresenceled to fewer
kin ties and smallernetworksoverall.
Net of othervariables,paid employmenthad
little impact on absolute networkcomposition
otherthanincreasingthe numberof ties to group
members, the unremarkableeffect of increas-
ing ties to coworkers,and the previouslynoted
decrease in kin ties among women. But the re-
lated variables of education, occupation, and
income had stronger effects, generally being
732 AMERICANSOCIOLOGICALREVIEW

Table 3. Regressionof Network Compositionon Genderand StructuralVariableswith InteractionTerms

# Kin # Kin Types # Advisors

Category/ Full Trimmed Full Trimmed Full Trimmed
IndependentVariables Model Model Model Model Model Model
Gender 1.295* 1.097* .862 .647 .090 .024
(2.00) (2.03) (1.96) (1.77) (.28) (.42)
Age
Age -.019 -.018 -.017 -.017 .013 .016
6
(-1.19) (-1.10) (-1.57) (-1.58) (1.68) (2.61)
Age2 .192 .165 .134 .127 -.122 -.166*
(1.14) (1.01) (1.18) (1.15) (-1.49) (-2.57)
Education .027 .018 .018 .009 .028** .026**
(1.32) (1.28) (1.26) (.93) (2.81) (3.70)
Employment
Full-timeemployment -.233 -.245 -.164 -.168 .051 -.080
(-1.57) (-1.67) (-1.63) (-1.68) (.70) (-1.54)
Part-timeemployment .217 .218 .146 .150 .118 .044
(1.40) (1.42) (1.40) (1.44) (1.57) (.73)
Occupation -.050 -.025 .001 .023 .090 .116-
(-.40) (-.21) (.01) (.29) (1.48) (2.65)
Income -.011 -.010 -.005 -.005 -.007* -.001
(-1.51) (-1.46) (-1.07) (-.97) (-1.99) (-.50)
Family
Maritalstatus .680** .654** .684** .691* -.067 -.082
(6.87) (8.49) (10.21) (13.26) (-1.39) (-1.74)
Children -.061 -.071* -.016 -.020 -.009 -.014
(-1.30) (-2.11) (-.49) (-.87) (-.38) (-.66)

Interactionof gender (male) with:

Age
Age -.063* -.067** -.047** -.046** .008
(-2.41) (-2.66) (-2.62) (-2.72) (.62)
Age2 .562* .612* .426* .434* -.119
(2.05) (2.33) (2.30) (2.45) (-.89)
Education -.017 -.017 -.006
(-.61) (-.86) (-.43)
Employment
Full-time employment .230 .244 .182 .192 -.255*
(1.02) (1.09) (1.19) (1.26) (-2.32)
Part-timeemployment -.354 -.350 -.291 -.293 -.182
(-1.33) (-1.32) (-1.62) (-1.63) (-1.41)
Occupation .085 .036 .101 .054 .062
(.47) (.22) (.82) (.49) (.70)
Income .010 .009 .006 .005 .008* --
(1.14) (1.04) (.98) (.83) (1.98)
Family
Maritalstatus -.063 .019 .125 .139
(-.39) (.17) (1.62) (1.89)
Children -.014 -.007 .043 .055
(-.21) (-.16) (1.30) (1.82)
Intercept 1.739** 1.853* 1.394"' 1.500** -.450* -.415**
(4.06) (4.79) (4.80) (5.73) (-2.16) (-2.61)
R2 .092 .091 .168 .167 .062 .056

F for interactionset 2.264 3.306 2.451 3.552 1.915 4.344

d.f. 9,1328 6,1331 9,1328 6,1331 9,1328 2,1335
Significance level .016 .003 .009 .002 .046 .013

sp <.05 **p <.01

Notes: Unstandardized coefficients, t-statistics are in parentheses. For variable descriptions see Table 2.
DETERMINANTSOF MEN'S AND WOMEN'S NETWORKS
positively associatedwith networksize andties
to nonkinoverall as well as to coworkers,advi-
sors, and friends.The numberof kin ties, how-
ever, was not dependenton education,income,
or occupation. The general pattern of these
variables indicates that, as expected, persons
who aremoreeconomically,educationally,and
occupationallyprivilegedformlargernetworks
and establishmore close ties to nonkinof vari-
ous types thando less privileged persons.
RelativeNetworkComposition
Broadly similar patterns were found for the
measures of relative network composition
(Tables 2 and 3). Genderwas a significantfac-
tor only for kin ties, with women's networks
being comprised of larger proportionsof kin
and more types of kin than were the men's.
There were also a few significant interaction
effects involving gender, but only in the equa-
tion for kin types. As before, these were with
employmentstatus(F41331 = 3.448,p < .01) and
age (F2 1331 = 4.458,p < *05).The role of gender
in the kin equationsindicates that women, es-
pecially those who are not employed full time,
not only maintainedclose ties to more kin, but
also maintainedmore diverse kin ties than did
similarly placed men. This result is consistent
with women's roles as "kinkeepers,"persons
who keep members of the extended family in
touchwith one another(Rosenthal1985, p. 965;
di Leonardo1987, p. 443).
Some social structuralvariableshad stronger
effects thangenderon relativenetworkcompo-
sition. Maritalstatushadthe strongestandmost
consistenteffect; being marriedled to a higher
proportionandgreatervarietyof kin types while
having the opposite effect on ties to nonkin.
The effects of education, employment status,
and occupation were also noteworthy for the
diversity of nonkin types. In addition, educa-
tion reduced the proportionof close ties in-
volving kin.
SUMMARY AND CONCLUSION
In the aggregate, women and men differed
considerably in their network compositions
thoughnot in networksizes. Comparedto men,
women's networkswere comprisedof morekin
and fewer nonkin (except neighbors). Most of
these gender differences disappearedor were
reduced, however, when structuralvariables
were controlled.In particular,men and women
733
had networks that contained similar numbers
of nonkin of various types when variables re-
lated to work, family, and age were controlled.
However, some gender differences remained,
primarily in ties to kin where women's net-
works contained more and larger proportions
of kin as well as more types of kin.
Contraryto expectations, gender and struc-
turalvariablesrarelyinteractedin their impact
on networkcomposition. Marriageand parent-
hood, for instance, did not impact negatively
on nonkin network formation for women.
However, a few gender-specific effects oc-
curred, particularly for employment-related
variables and age. Full-time employment de-
creasedthe numberand diversityof kin only in
women's networks.Also, increasingage had a
strongerimpacton men's thanon women's kin
ties. Severalstructuralvariablesconnectedwith
employment,maritalstatus,andage were more
importantthan gender in their impact on net-
workcomposition.Ties to nonkin,for instance,
were less numerousamong marriedand older
persons,butmorecommonamongpersonswho
were highly educated or in professional and
managerialoccupations.The structuralvariables
often had the opposite effects for ties to kin.
Overall, these analyses offer considerable
support for the structuralperspective. Net of
othervariables,genderhad little impacton the
absoluteandrelativenonkinmeasures,indicat-
ing that men and women with similar family-
and work-relatedcharacteristicshave nonkin
networksthat are also similar.
On the other hand, structuralvariables did
not fully eliminate the effect of gender on kin
ties. In theirpersonalnetworks,womenincluded
more and larger proportionsof kin as well as
more diverse kin types than did similarly situ-
atedmen, althoughthe disparitieswere reduced
to some degree when women workedfull-time.
Women may be disposed to focus more of their
close ties on family members, men more on
ties to nonkin.These resultsareconsistentwith
studies of caregiving, which typically have
found that women are expected to be and are
more active thanmen in the maintenanceof kin
ties and in the care of dependentrelatives, such
as an aged parent(e.g., Leigh 1982; Rosenthal
1985; Sherman,Ward,and LaGory 1988). But
a structuralfactor, proximity,-may also be im-
portant:Older women have been found to live
nearerto theirgeographicallyclosest child than
do men of the sameage (SpitzeandLogan 1989,
Table 3). The proximity factor may also be
734
importantto younger women (perhapsin rela-
tion to theirparents);it could offer the opportu-
nity to maintainstrong ties to relatives, since,
in spite of the telephone, frequencyof kin in-
teractionis often foundto dependmost strongly
on geographicdistance(Lee 1980,pp. 928-929).
In balance, the evidence here supports the
structuralperspective:Most genderdifferences
in networkswere due to opportunitiesand con-
straintsarising out of women's and men's dif-
ferent locations in the social structure.Men's
and women's networks differed more as a re-
sult of thatfact and less because they were pre-
disposed to form and maintain differing net-
works.Blau's work is importantin understand-
ing that networks form within an opportunity
context that facilitates or impedes the forma-
tion of social ties (1977; see also Marsdenforth-
coming). The social structureshapes opportu-
nitiesto formrelationshipsof varioussorts(Blau
1977). If women do not work for pay, they
cannot have coworkers in their personal net-
works. Unmarriedpersons cannot form ties to
spouses or in-laws. These structuralfactors
provide the context within which personalties
form.
The structuralperspective has implications
beyond the realm of the personalnetworksin-
vestigated here. Kanter (1977) used it to ex-
plain gender differences in networks and be-
havior in organizations.She contended that it
is not the dispositions of men and women, but
the small numberof women in high-level posi-
tions and theirrelativelack of opportunityand
power in male-dominated organizations that
explain the higherperformanceandmobility of
men in workorganizations.Largernumbersand
proportionsof women in high-level organiza-
tional positions would decrease women's dis-
advantage.Kanter'swork on genderin organi-
zations has pointed to the power of structural
effects on behaviorthatwas previouslyascribed
to individualpredispositions.
This examinationof strongnetworkties con-
firms previous findings of gender differences
in networks while also illuminatingtheir ori-
gins. These resultssupportBlau's andKanter's
conclusions:If men and women were in similar
social structuralpositions theirbehaviorwould
differ little. As more women move into paid
employment, the genders' network composi-
tions can be expected to become more alike,
with more close ties to nonkin, especially co-
workers, and fewer ties to kin. Women may
still maintaina largernumberof ties to kin than
AMERICANSOCIOLOGICALREVIEW
do men, however, as long as they remain the
primary caretakers and kinkeepers in most
families.
This research also offers clear support for
Epstein's more general claim, that "the over-
whelmingevidence createdby the past decades
of researchon gender supportsthe theory that
genderdifferentiation... is best explainedas a
social constructionrooted in hierarchy,not in
biology or in internalization. . ."(1988, p. 15).
GWEN MOORE, Assistant Professor of Sociology at
SUNY-Albany,is currentlyfocusingher research on
the intersections of gender, networks, and power.
This research includes the study of cross-national
variation in the relationshipof gender and author-
ity, and an analysis of gender and managerial net-
works in the New YorkState government."EliteIn-
tegration in Stable Democracies: A Reconsidera-
tion" (forthcoming,EuropeanSociological Review),
an article written with John Higley, Ursula
Hoffmann-Lange,and Charles Kadushin, summa-
rizes more than a decade of research on the struc-
ture of elite networks.WithJ. Allen Whitt,Moore is
co-editingThe PoliticalConsequencesof Social Net-
works.
REFERENCES
Aukett,Richard,JaneRitchieandKathrynMill. 1988.
"GenderDifferences in FriendshipPatterns."Sex
Roles 19:57-66.
Berk,SarahFenstermaker.1988. "Women'sUnpaid
Labor: Home and Community."Pp. 287-302 in
WomenWorking:Theoriesand Facts in Perspec-
tive, 2nd. edition,editedby Ann HeltonStromberg
and Shirley Harkess.MountainView, CA: May-
field.
Blau, Peter M. 1977. Inequalityand Heterogeneity:
A PrimitiveTheoryof Social Structure.New York:
Free Press.
Brass, Daniel J. 1985. "Men's and Women's Net-
works:A Study of InteractionPatternsand Influ-
ence in an Organization."Academy of Manage-
mentJournal 28:327-43.
Burt, Ronald S. 1983. "Range."Pp. 176-94 in Ap-
plied NetworkAnalysis:A MethodologicalIntro-
duction,by Ronald S. Burt,Michael J. Minorand
Associates. Beverly Hills, CA: Sage.
._ _ 1985. "General Social Survey Network
Items."Connections.8:119-23.
Caldwell, MaytaA. and Letitia Anne Peplau. 1982.
"Sex Differences in Same-Sex Friendship."Sex
Roles 8:721-32.
Campbell, Karen E. 1988. "GenderDifferences in
Job-RelatedNetworks." Workand Occupations
15:179-200.
Campbell, Karen E., Peter V. Marsdenand Jeanne
S. Hurlbert.1986. "Social Resources and Socio-
DETERMINANTSOF MEN'S AND WOMEN'S NETWORKS 735

economic Status."Social Networks8:97-117. the Family Life Span."Journal of Marriage and

Chodorow,Nancy. 1978. TheReproductionof Moth- the Family 44:197-208.
ering. Berkeley: University of California. Lin, Nan. 1982. "Social Resourcesand Instrumental
Davis, James A. and Tom W. Smith. 1988. General Action." Pp. 131-46 in Social Structureand Net-
Social Surveys, 1972-1988 [MRDF]. Principal work Analysis, edited by Peter V. Marsden and
Investigator,James A. Davis; Director and Co- Nan Lin. Beverly Hills, CA: Sage.
PrincipalInvestigator,Tom W. Smith. NORC ed. Lin, Nan, WalterM. Ensel, and J. C. Vaughn. 1981.
Chicago:NationalOpinionResearchCenter[pro- "Social Resources and Strength of Ties: Struc-
ducer]. Storrs, CT: The Roper Center for Public turalFactorsin OccupationalStatusAttainment."
OpinionResearch,Universityof Connecticut[dis- AmericanSociological Review 46:393-405.
tributor].1 datafile (23,356 logical records)and1 Lipset, Seymour Martin, Martin Trow and James
codebook (790 pp.). Coleman. 1956. Union Democracy. New York:
di Leonardo, Michaela. 1987. "The Female World Doubleday.
of Cardsand Holidays:Women, Families and the Marsden,PeterV. 1987. "CoreDiscussion Networks
Work of Kinship."Signs: Journal of Women in of Americans." American Sociological Review
Cultureand Society 12:440-58. 52:122-31.
Epstein, Cynthia Fuchs. 1988. Deceptive Distinc- . Forthcoming. "Network Diversity, Sub-
tions: Sex, Gender and the Social Order. New structures, and Opportunities for Contact." in
Haven:Yale University. Structures of Power and Constraint: Essays in
Fischer, Claude. 1982. To Dwell Among Friends. Honor of Peter M. Blau, editedby CraigCalhoun,
Chicago:University of Chicago MarshallW. Meyer and W. Richard Scott. New
Fischer, Claude, and Stacey Oliker. 1983. "A Re- York: Cambridge.
search Note on Friendship,Gender, and the Life Marsden, Peter V., and Jeanne S. Hurlbert. 1987.
Cycle." Social Forces 62:124-132. "SmallNetworksandSelectivityBias in theAnaly-
Gerstel,Naomi. 1988. "Divorce, Genderand Social sis of Survey Network Data." Social Networks
Integration."Genderand Society 2:343-67. 9:333-49.
Gilligan, Carol. 1982. In a Different Voice: Psycho- Marsden,PeterV., andNan Lin. 1982. Social Struc-
logical Theoryand Women'sDevelopment.Cam- ture and Network Analysis. Beverly Hills, CA:
bridge, MA: Harvard. Sage.
Granovetter,MarkS. 1974. Getting a Job: A Study McPherson,J. Miller and Lynn Smith-Lovin. 1982.
of Contacts and Careers. Cambridge,MA: Har- "Women and Weak Ties: Differences by Sex in
vard. the Size of VoluntaryOrganizations."American
. 1982. "The Strength of Weak Ties: A Journal of Sociology 87:883-904.
NetworkTheory Revisited."Pp. 105-30 in Social . 1986. "Sex Segregation in Voluntary
Structuresand NetworkAnalysis, edited by Peter Associations."American Sociological Review
V. MarsdenandNan Lin. BeverlyHills, CA: Sage. 51:61-79.
Hurlbert,Jeanne S.and Alan C.-Acock. 1990. "The Miller, Jean Baker. 1976. Toward a New Psychol-
Effects of MaritalStatuson the FormandCompo- ogy of Women.Boston: Beacon.
sition of Social Networks."Social Science Quar- Miller, Jon. 1986. Pathways in the Workplace.
terly 71:163-74. Cambridge:Cambridge.
Kadushin,Charles.1982. "Social Density and Men- Moore, Gwen. 1979. "The Structureof a National
tal Health." Pp. 147-58 in Social Structureand Elite Network." American Sociological Review
NetworkAnalysis,editedby PeterV. Marsdenand 44:672-92.
Nan Lin. Beverly Hills, CA: Sage. Rosenthal, Carolyn J. 1985. "Kinkeeping in the
Kanter,Rosabeth Moss. 1977. Men and Womenof FamilialDivision of Labor."Journal of Marriage
the Corporation.New York: Basic.
and the Family 47:965-74.
Laumann, Edward 0. and Franz U. Pappi. 1976.
Susan, Russell Ward and Mark LaGory.
Networksof Collective Action: A Perspective on Sherman,
CommunityInfluence Systems. New York: Aca- 1988. "Women as Caregiversof the Elderly: In-
demic. strumentaland ExpressiveSupport."Social Work
Laumann,Edward0. and David Knoke. 1987. The 33:164-67.
OrganizationalState: Social Choice in National Spitze, Glennaand JohnLogan. 1989. "GenderDif-
Policy Domains. Madison, WI: University of ferences in Family Support:Is There a Payoff?"
Wisconsin. The Gerontologist29:108-13.
Lee, Gary R. 1980. "Kinship in the Seventies: A Wellman,Barry.1985. "DomesticWork,PaidWork
Decade Review of Research and Theory."Jour- and Net Work." Pp. 159-91 in Understanding
nal of Marriage and the Family 42:923-34. Personal Relationships,editedby Steve Duck and
Leigh, Geoffrey K. 1982. "KinshipInteractionover Daniel Perlman.London:Sage.