Sibling Relationships Over the Life Course: A Panel Analysis

14
Journal of Marriage and Family 63 (May 2001): 555–568 555 LYNN WHITE University of Nebraska—Lincoln l Sibling Relationships Over the Life Course: A Panel Analysis Using pooled time series analysis on approxi- mately 9,000 individuals ages 16–85 interviewed in the 1987–1988 and 1992–1994 waves of the National Survey of Families and Households (NSFH), this research examines change in 4 be- havioral measures of sibling relationships—prox- imity, contact, giving help, and receiving help— over the life course. All four measures of sibling relationship decline significantly during early adulthood, but proximity and contact stabilize in middle age and do not decline further, whereas sibling exchange demonstrates a slight rise after approximately age 70. Life course analyses pro- vide only modest support for a model in which siblings substitute for parents, spouses, and chil- dren. With the partial exception of proximity, measured life course changes do not explain ob- served age effects. In the daily round of activities that occupy the foreground in adult lives, siblings generally play a minor role (Wellman & Wortley, 1990). If one shifts one’s perspective from a snapshot of daily life to a lifetime perspective, however, siblings play a more prominent role. Siblings are those with whom one most closely shares genetic, fam- ily, social class, and historical background and to whom one is tied for a lifetime by a network of interlocking family relationships. Cross-sectional evidence suggests that the average adult has con- Department of Sociology, University of Nebraska—Lin- coln, Lincoln, NE 68588-0324 ([email protected]). Key Words: life course, siblings. tact with a sibling once or twice a month for 60 or 70 years after leaving home (White & Ried- mann, 1992), and most adults perceive significant reciprocal obligations between siblings (Connidis, 1994; Rossi & Rossi, 1990). Although the sibling relationship may not be central in most adults’ lives at any given point in time, it is unique in its durability. In a pattern familiar to other areas of family study, research on sibling relationships tends to focus either on childhood or old age. Although developmentalists stress the importance of sib- lings as childhood rivals and mentors (Hethering- ton, 1994) and gerontologists argue that siblings take on a renewed importance in old age (Gold, 1989), researchers know little about sibling rela- tionships between 20 and 60. Almost all of the research on siblings is based on small, nonrandom samples in a cross-sectional or retrospective de- sign. Even for the most often-cited conclusions, such as the increased importance of siblings in old age, the research base for adult sibling relation- ships is weak. The present study uses pooled time series techniques on a large national panel sam- pled by the National Survey of Families and Households (NSFH) in 1987–1988 and 1992– 1994 to extend the knowledge of how sibling re- lationships change over time. Using four measures of sibling relationships (proximity, contact, giving help, and receiving help), it describes changed sib- ling relationships from age 18 to 85 and examines whether observed changes are attributable to re- lated family life course changes, such as parental and marital status. The data set does not contain

Transcript of Sibling Relationships Over the Life Course: A Panel Analysis

Page 1: Sibling Relationships Over the Life Course: A Panel Analysis

Journal of Marriage and Family 63 (May 2001): 555–568 555

LYNN WHITE University of Nebraska—Lincoln

l

Sibling Relationships Over the Life Course:

A Panel Analysis

Using pooled time series analysis on approxi-mately 9,000 individuals ages 16–85 interviewedin the 1987–1988 and 1992–1994 waves of theNational Survey of Families and Households(NSFH), this research examines change in 4 be-havioral measures of sibling relationships—prox-imity, contact, giving help, and receiving help—over the life course. All four measures of siblingrelationship decline significantly during earlyadulthood, but proximity and contact stabilize inmiddle age and do not decline further, whereassibling exchange demonstrates a slight rise afterapproximately age 70. Life course analyses pro-vide only modest support for a model in whichsiblings substitute for parents, spouses, and chil-dren. With the partial exception of proximity,measured life course changes do not explain ob-served age effects.

In the daily round of activities that occupy theforeground in adult lives, siblings generally playa minor role (Wellman & Wortley, 1990). If oneshifts one’s perspective from a snapshot of dailylife to a lifetime perspective, however, siblingsplay a more prominent role. Siblings are thosewith whom one most closely shares genetic, fam-ily, social class, and historical background and towhom one is tied for a lifetime by a network ofinterlocking family relationships. Cross-sectionalevidence suggests that the average adult has con-

Department of Sociology, University of Nebraska—Lin-coln, Lincoln, NE 68588-0324 ([email protected]).

Key Words: life course, siblings.

tact with a sibling once or twice a month for 60or 70 years after leaving home (White & Ried-mann, 1992), and most adults perceive significantreciprocal obligations between siblings (Connidis,1994; Rossi & Rossi, 1990). Although the siblingrelationship may not be central in most adults’lives at any given point in time, it is unique in itsdurability.

In a pattern familiar to other areas of familystudy, research on sibling relationships tends tofocus either on childhood or old age. Althoughdevelopmentalists stress the importance of sib-lings as childhood rivals and mentors (Hethering-ton, 1994) and gerontologists argue that siblingstake on a renewed importance in old age (Gold,1989), researchers know little about sibling rela-tionships between 20 and 60. Almost all of theresearch on siblings is based on small, nonrandomsamples in a cross-sectional or retrospective de-sign. Even for the most often-cited conclusions,such as the increased importance of siblings in oldage, the research base for adult sibling relation-ships is weak. The present study uses pooled timeseries techniques on a large national panel sam-pled by the National Survey of Families andHouseholds (NSFH) in 1987–1988 and 1992–1994 to extend the knowledge of how sibling re-lationships change over time. Using four measuresof sibling relationships (proximity, contact, givinghelp, and receiving help), it describes changed sib-ling relationships from age 18 to 85 and examineswhether observed changes are attributable to re-lated family life course changes, such as parentaland marital status. The data set does not contain

Page 2: Sibling Relationships Over the Life Course: A Panel Analysis

556 Journal of Marriage and Family

measures of affection or psychological importanceof the sibling relationship, however the analysisprovides useful descriptive information aboutchanges in the behavioral dimensions of siblingrelationships over a broad age range.

BACKGROUND LITERATURE

Conceptual Frameworks

Underlying most interpretations of sibling rela-tionships is the assumption of a hierarchy of kin-ship relationships. This standard kinship modelenvisions family ties as a set of nested circles. Theinner circle contains those with whom one has theclosest relationships, those who have the strongestclaims on one another. The second and outer ringscontain those with progressively less demandingattachments. Although siblings are considered tobe members of the inner circle during childhood,it is normative for adults to reorganize their innercircles to give children, spouses, and parents pri-ority over siblings (Parsons, 1943). Rossi andRossi’s (1990) research on normative obligationsto various classes of kin suggests that this normis widely shared. Although the average adult feelsa substantial obligation to help a sibling in timeof need and to share celebratory occasions withsiblings, this obligation is much lower than thatreported for parents and children and approxi-mately on a par with that felt toward grandparents,stepchildren, or stepparents. Both developmentaland life course perspectives offer frameworks forunderstanding how the hierarchy of family obli-gations changes over time.

Developmental perspectives. Along with parents,siblings form critical reference points in the de-velopment of identity and attachment in earlychildhood (Hetherington, 1994). Although verylittle research or theory has considered the siblingrelationship during the transition to adulthood(Dunn, 1984), general development perspectivesprovide some guidance. The chief developmentaltask of young adulthood is individuation, the de-velopment of psychological and instrumental au-tonomy from one’s parents (Bowen, 1976; Erik-son, 1959). The child reorganizes the parent-childrelationship on the basis of mutual respect, lettinggo of childhood rivalries and resentments, forgo-ing excessive feelings of either obligation or de-pendence, and becoming self-directed (Harvey &Bray, 1991). Although not explicit in theoreticalstatements, it seems likely that development of

personal autonomy requires resolving sibling ri-valries, resentments, and dependencies as well.Maturity does not mean absence of close ties tothe family of origin, though it does imply thatthose who resolve this developmental stage suc-cessfully are free to form new attachments.

At the other end of the life course, gerontolo-gists have suggested that sibling relationships mayagain become salient as elders seek out those in-dividuals with whom they share the longest his-tory to engage in life review. Among those whostill carry emotional baggage from childhood, thisis a final chance to resolve rivalries and compe-tition (Goetting, 1986).

In developmental perspectives, age itself is afactor affecting identities and relationships. Thisperspective suggests that sibling relationshipsshould become less important as individuals makethe transition to adulthood and perhaps becomemore important again in later life.

Life course perspectives. Instead of focusing onthe unfolding of intrinsic age-related develop-ment, life course perspectives suggest that the ag-ing experience may be analyzed by consideringthe series of life course transitions that actuallystructure individual movement across time (Hohn& Mackensen, 1989). Rather than treating homeleaving, marriage, and childbearing as normativecomponents of human development, life coursetheorists treat these as variables that can be puttogether in various ways to create an individuallife course. In the case of siblings, this perspectivesuggests that if and when individuals get married,have children, or establish careers, sibling rela-tionships become less central. This is not seen asa developmental corollary of aging, however, butof specific life course transitions. In the absenceof these transitions, i.e., among those who nevermarry or never have children or never leave home,this perspective suggests that siblings should re-main important.

One implication of the life course perspectiveis that if these transitions are reversed (i.e., if in-dividuals become widowed or divorced or whentheir parents die), then siblings should becomemore central parts of their social network. Thelatter situation is captured in Cantor’s (1979) hi-erarchical-compensatory model, which suggeststhat siblings are first pushed out of the inner circleto make room for spouses and children and thenpulled back if deficits arise among these preferredkin. Cantor’s model clearly suggests that siblingsare second-best members of adult kin networks;

Page 3: Sibling Relationships Over the Life Course: A Panel Analysis

557Sibling Relationships

however, the same set of hypotheses—that sib-lings become more important after losses tospouses, children, or parents—is also consistentwith less value-laden models that stress opportu-nity. More generally, then, the family life coursemodel of sibling relationships suggests that sib-lings are permanent but flexible members of oursocial networks, whose roles in our networks arerenegotiated in light of changing circumstancesand competing obligations.

Critique. Both the developmental and the lifecourse perspectives of sibling relationships areconsistent with the notion that a normative hier-archy of family relationships cuts across multipledomains. Several gerontologists have challengedthis notion. Specifically referring to the social net-works of the elderly, Connidis and Davies (1990,1992) and Simons (1983–1984) argue that the hi-erarchy of kin preference depends on the task. Forsome tasks, siblings or friends might be preferredto spouses or children, even if the latter are avail-able. Further, consistent with Elder’s (1974) ar-guments, they argue that relationships evolve overtime and that historical roots as well as contem-porary circumstances affect whether siblings, forexample, are second-best or preferred members ofnetworks. Such arguments suggest that resultsmight differ across dimensions of the sibling re-lationship and that sustaining sibling relationshipsmight not correlate as neatly with changes in thefamily life course as the general hierchical modelsuggests.

In some ways, the developmental and the lifecourse perspective predict the same general pat-tern of sibling relationships—declining sibling re-lationships in young adulthood and an increase inold age—but they attribute it to different under-lying processes and with differential degrees ofdeterminism. The other important difference be-tween them is their implicit dependent variable.Developmental theory stresses emotional ties suchas attachment and meaningfulness (Avioli, 1989;Cicirelli, 1985), whereas the life course model in-cludes a broader range of instrumental and ex-pressive dimensions. Because this research is lim-ited to behavioral measures of siblingrelationships, it is less likely to find the upturn inelders’ sibling relationships that developmentaltheory predicts.

Prior Empirical Research

Research on the relationship between sibling tiesand age is uniformly cross-sectional and surpris-

ingly sparse. The few studies that examine asso-ciations between age and behavioral measuressuch as contact and exchange generally reportnegative relationships. An inverse relationship be-tween age and sibling contact is reported by Ro-senberg and Anspach (1973) among a sample ofWhite, working-class Philadelphia residents age45–79 and by White and Riedmann (1992) for alarge national sample age 18–95 drawn from thefirst wave of NSFH. An inverse relationship be-tween age and measures of actual exchange in theNSFH sample is reported by White and Riedmann(1992) and also, focusing only on those over age55 with geographically close kin, by Miner andUhlenberg (1997). Suggs (1985) also reported anegative association between age and helping ina small sample of African American elders. Con-nidis and Campbell (1995) found no associationbetween age and contact in their over-55 samplebut this may be because they controlled for emo-tional as well as physical closeness. If we shift toemotional rather than behavioral measures, thenseveral cross-sectional studies report a positivecorrelation between age and sibling attachmentamong older populations (Connidis & Campbell,1995; Lopata, 1973); White and Riedmann, 1992).

Predictions regarding family life course tran-sitions find general support from cross-sectionalresearch. Using a sample ranging from age 18 to95, White and Riedmann (1992) reported that sib-ling exchange and affection were significantlylower for married respondents and that contact,perceived support, and exchange were significant-ly lower for respondents who had adult children.Using the same data set but focusing on respon-dents over 55 with nearby siblings, Miner andUhlenberg (1997) reported substantially similarfindings. Johnson and Catalano (1981) also re-ported that sibling interaction was less importantand less frequent among elderly with children thanamong childless elderly. In a retrospective studyof 60 dyads, Connidis (1992) found that, amongthe approximately half who reported a change insibling relationships following life course transi-tions, almost all reported that these transitions—having children as well as becoming unmarried—were associated with stronger sibling relation-ships.

The Present Study

Theory and empirical generalizations drawn fromcross-sectional research suggest that relationshipswith siblings change over the years of adulthood

Page 4: Sibling Relationships Over the Life Course: A Panel Analysis

558 Journal of Marriage and Family

because of age or life course events. None of theseinferences, however, has been supported yet bylongitudinal analysis. This research uses data ona sample of nearly 9,000 adults who were askedcomparable questions about sibling relationshipsin two interviews 4–7 years apart. By piecing to-gether changes reported by adults at all ages from18 to 85, it is possible to estimate changes in ex-change and contact among siblings over the lifecourse. I test the hypothesis that sibling relation-ships have a curvilinear relationship with age, de-clining in young adulthood and then resurfacingin old age. Further, I examine the following lifecourse changes: becoming married, becoming un-married (widowed or divorced), adding children,becoming geographically closer to siblings, be-coming orphaned, and changing sibling number.If the age effect is reduced by control for familylife course measures, then this will suggest thatobserved age effects are attributable to new re-sponsibilities rather than a developmental process.This analysis is viewed better as a description thanas a test between two theories, however. Not onlydoes the analysis lack a measure of the meaning-fulness of the sibling tie, as demanded by devel-opmental theory, but it also includes only selec-tive measures of the respondent’s family lifecourse and includes no information at all on thesiblings’ circumstances.

The analysis examines four behavioral mea-sures of sibling relationships—proximity, contact,giving help, and receiving help. Although prox-imity is a resource for sibling interaction (i.e., apredictor of contact and exchange), proximity canalso be a deliberate choice that reflects importanceof family ties. Following the typology of Bengt-son, Vern, and Roberts (1991), geographical prox-imity is included as a measure of the strength ofthe sibling tie but also introduced as an interven-ing and moderating variable in other analyses.

This research controls three characteristics ofthe sibship that are related to exchanges amongadult sibs in cross-sectional research: gender com-position, number, and physical proximity. Sub-stantial research finds that sisters have strongerrelationships than either brother-brother or broth-er-sister combinations (Connidis, 1989; Lee, Man-cini, & Maxwell, 1990; White & Riedmann, 1992)and that geographical proximity is important in allmeasures of sibling relations (Connidis, 1989; Leeet al., 1990; Suggs, 1985; White & Riedmann,1992). Whether because larger sibships are actu-ally emotionally closer or simply because a largernumber increases the odds that one sibling will be

proximate and compatible, a larger sibship is cor-related with greater exchange, affection, contact,and perceived support (White & Riedmann,1992). The data do not permit measurement of agedifferences among siblings.

Race and ethnicity show cross-sectional asso-ciations with measures of sibling relationshipstrength. Treas and Bengtson (1987) concludedthat Hispanics are more likely than either Anglosor African-Americans to report mutual affectionand exchange among siblings. Taylor, Chatters,and Mays (1988) concluded that African Ameri-cans are less likely than Anglos or Hispanics torely on siblings for support in an emergency. Pre-vious research from the 1987–1988 NSFH cross-section (Riedmann & White, 1996) showed thatthe direction of race-ethnic differences dependedon how sibling relationships were measured, withno group showing consistently stronger sibling re-lationships. The latter research did find that direc-tion of effects for Latinos was sometimes reversedwhen controlling for immigrant status, suggestingthat this is an important control variable.

Extended family ties have been reported to beless important as social class rises (Fischer, 1982;Treas & Bengtson, 1987), and a small study byAllan (1977) showed that more working- thanmiddle-class respondents describe a sibling as abest friend. Nevertheless, the most common em-pirical pattern is for family obligations generally(Rossi & Rossi, 1990) and sibling exchange inparticular (Miner & Uhlenberg, 1997; White &Riedmann, 1992) to be stronger among those withhigher educations and incomes.

Women are the family ‘‘kin-keepers’’ (Hages-tad, 1986), and research demonstrates that womenare more likely to initiate and maintain ties withkin, including siblings (Fischer 1982; Lee et al.,1990; Reiss & Oliveri, 1983; Treas & Bengtson,1987; White & Riedmann, 1992). Whether wom-en are more apt to strengthen these ties over timeis unknown.

Analytic Technique

The primary analytic technique for this analysis israndom effects pooled time series analysis that al-lows us to estimate the effect of characteristicsthat vary within as well as across individuals, forexample age and family life course. The depen-dent variable in these analyses is the first differ-ence equations in changes in sibling relationshipsbetween waves regressed on changes in age andfamily life course, with controls for time-invariant

Page 5: Sibling Relationships Over the Life Course: A Panel Analysis

559Sibling Relationships

characteristics of individuals such as gender, race,and ethnicity. The strengths of the random effectssolution are its statistical power and its ability toassess the effect of time-invariant characteristics(such as gender) on change in the dependent var-iable (Allison, 1994). With a fully-specified mod-el, the random effects solution is both efficient andappropriate. On the other hand, if the model omitsimportant time-invariant individual characteristics,a random effects solution may be biased. Becauseindividual norms of family obligations are notmeasured or controlled in this analysis, it is pos-sible that variance more properly attributed todeep seated values and attitudes—values and at-titudes that may vary systematically with cohort—is attributed to age or family life course transi-tions. To overcome this problem, I estimate a sup-plemental fixed effects solution to assess the ro-bustness of age and life course effects.

Method

Sample. The sample of adults with siblings isdrawn from the NSFH panel. The first wave in-terviewed a random sample of 13,017 adults iden-tified through an area probability process. Thesampling design oversampled African Americans,Latinos, Puerto Ricans, the recently married, co-habitors, and those in stepfamilies (Sweet, Bum-pass, & Call, 1988). The response rate was 74%.In 1992–1994, reinterviews were completed with10,008 or 76.9% of the original respondents. At-trition was greater for those under 25 and over 75,for males (especially those never married, di-vorced, and separated), for cohabitors of both sex-es, and for nonAnglos. This study uses a subsam-ple of individuals who reported a living sibling inboth waves. To reduce the likelihood that age ef-fects would be distorted by small numbers at ad-vanced ages, I omitted those over 85 (n 5 95).The analysis is based on 8,838 individuals age 18–85 in 1987 who reported a sibling in both waves.Because the sample is so large, minimal effectswith little substantive importance reach the .05level. To reduce the likelihood of attributing sub-stantive meaning to chance effects, I consider onlyfindings that reach the .01 level of significance.

The complex sampling design used in this pro-ject included both substantial clustering and ov-ersampling. Because many of the variables usedto select the oversample are included as controlvariables, these analysis do not weight the sample(Winship & Radbill, 1994). The clustered sam-pling design is potentially more problematic as it

may result in underestimation of the actual errorvariance in the coefficients (Lee, Forthofer, &Lorimer, 1989). Balanced repeated replicate tech-niques were used to estimate the extent to whichthe standard errors provided by the computingsoftware (STATA) were underestimated. Designeffects were found to be very small (rarely morethan 1.1, available from author) and did not affecteither the statistical significance of individual co-efficients or substantive interpretations. As a re-sult, the standard errors provided by the standardcomputing package are reported throughout.

Measures

Dependent variables. Four indicators of sibling re-lationships are used: proximity, contact, givingsupport, and receiving support. The Time 1 seriesasked separately about step- and full-siblings,whereas the Time 2 series combined them. Mea-sures for Time 1 use data about relationships withfull siblings except for those 366 individuals whoonly had stepsiblings.

The data set does not provide actual distanceof siblings but it includes the number of siblingsliving less than 2 miles, between 2 and 25 miles,between 25 and 300 miles, and more than 300miles away. Following Miner & Uhlenberg(1997), proximity is a dummy variable coded 1 ifrespondent had any siblings within 25 miles. Thisdummy variable violates the assumptions of acontinuous dependent variable specified forpooled time series analyses, but more complexmanipulations of these sketchy data (e.g., assign-ing a midpoint to each interval) and alternativeanalysis strategies (e.g., logistic regression usinga linear panel approach) produced very similar re-sults. To retain comparability with Miner and Uhl-enberg and to provide a similar format to the othervariables, I report results for the dummy measure.In each wave, 58% of the sample were within 25miles of a sibling but this similarity masks indi-vidual shifts. Sixteen percent were newly within25 miles of a sibling at Time 2.

Sibling contact is measured by the sum of twoquestions, each having a set of response categoriesranging from 1 (no contact) to 6 (more than oncea week): ‘‘During the past 12 months, how oftendid you see any of your [full] brothers or sisters?’’and ‘‘During the past 12 months, how often didyou talk on the telephone or receive a letter fromany of your [full] brothers or sisters?’’ Mean sib-ling contact was 7.89 (SE 5 .72) at time 1 and7.82 (SE 5 .58) at Time 2. These scores represent

Page 6: Sibling Relationships Over the Life Course: A Panel Analysis

560 Journal of Marriage and Family

levels of sibling contact approximately equivalentto contact once or twice a month.

Exchange is measured by two variables, eachof which is a count of ‘‘yes’’ or ‘‘no’’ items tap-ping whether respondent gave to or received fromsiblings. Respondents were asked whether theyhad given or received the following six kinds ofhelp: transportation, home repairs, housework, ad-vice/moral support, gifts of $200, or loans of$200. The financial questions covered the last 5years, and the others covered the month before theinterview. Respondents were specifically askedabout brothers and sisters as sources of and recip-ients of support. Levels of exchange were quitelow in both waves, with a mean of .72 forms ofaid given and .46 received in Wave 1 and .58 giv-en and .42 received in Wave 2.

Independent variables. Age is the primary inde-pendent variable necessary to assess the devel-opmental hypothesis. It is included as a continu-ous variable and also as a polynomial to examineproposed curvilinear effects. Because age cannotattain a true zero in this analysis, age is trans-formed to reflect the adult life span by subtracting18. Life course changes are tapped by the differ-ences between a series of dummy variables. Foreach wave, parents is scored 1 if respondent hasany living biological parents and 0 if both parentsare dead. The effects of marital change are as-sessed through two dummy variables, one mark-ing individuals who went from unmarried (nevermarried, separated, widowed, or divorced) to mar-ried between waves and the other marking thosewho went from married to unmarried. Those withno change in marital status are the omitted cate-gory in the change analysis. The number of re-spondent’s children in each wave includes chil-dren of all ages, step and biological, whether ornot living in the household.

Measuring change in number of siblingsproved to be a more complex task than anticipat-ed, with approximately equal numbers of respon-dents (15%) indicating sibling gain as sibling loss.Further analysis demonstrated that sibling gainwas associated with complex and changing familystructures as well as a change in question wording.As anticipated, greater age was the best predictorof sibling loss (White, 1998), and 50% of the re-spondents over age 70 reported the loss of at leastone sibling between waves. Sibling loss is indi-cated by a dummy variable scored 1 when therespondent reported fewer siblings at Time 2 thanTime 1, and another dummy variable marks cases

where the respondent reported more siblings atTime 2.

Control variables. The control variables are in-tended to adjust for time-invariant backgroundfactors that may affect sibling relationships, andTime 1 measures are used for all control variables.Dummy variables indicate whether respondent isfemale, African American, or Latino/Latina. Ini-tial size of the sibship ranges from 1 to 20 (M 53.6. Respondent’s education is measured contin-uously in years of school. Immigrant is coded 1for respondents who were born outside of theUnited States and 0 for native born.

RESULTS

Random Effects Analysis: Change in SiblingRelationships

The focus of these analyses is aging and lifecourse change in sibling relationships. For eachdependent variable, an initial model estimates ageeffects on changing sibling relationships control-ling for invariant background measures, and a sec-ond model tests whether the addition of life coursechanges explains the observed age effects. For thecontact and exchange measures, a third modelcontrols for the effect of changing proximity. Therandom effects analyses are summarized in Table1 for proximity and contact and in Table 2 for theexchange measures. Parameters in the random ef-fects solution are estimated from a covariance ma-trix in which the covariances between and withinindividuals are weighted to reflect their respectivevariance components, and the resulting coeffi-cients can be interpreted relative to the originalscale of measurement rather than to a changescore. Because effects are judged relative to theaverage score, a positive coefficient for femaleson contact, for example, may be interpreted assaying that women increase sibling contact morethan men or that they are more likely than men tosustain sibling contact. Because none of the mea-sures of sibling relationship in this analysis showmean increases, the latter interpretation is gener-ally more appropriate

Changing proximity. Sibling proximity declinesmodestly but significantly and nonlinearly withage. The likelihood of living within 25 miles of asibling declines from about 75% at age 18 toabout 55% at age 50 and then flattens out in mid-dle age (Model 1a). Those with more education

Page 7: Sibling Relationships Over the Life Course: A Panel Analysis

561Sibling Relationships

TABLE 1. CHANGING PROXIMITY TO SIBLINGS AND CHANGING CONTACT: RANDOM EFFECTS MODELS

Variable

Changed Proximity

1a 1b

Changed Contact

2a 2b 2c

Age

Age2 (310)

Age3 (31000)

Education

Immigrant

No. siblings

African American

Latino/Latina

Female

2.008**(.001).0006**

(.0001)ns

2.036**(.002)

2.249**(.021).026*

(.002).100**(.012).075**

(.020)2.016(.009)

2.009**(.001).0007**

(.0002)ns

2.036**(.002)

2.248**(.021).026**

(.002).100*

(.012).075**

(.020)2.016(.009)

2.131**(.012).024**

(.004)2.014**(.004)

2.036**(.009)

2.971**(.111).130**

(.009).746**

(.066).494*

(.107).476**

(.049)

2.135**(.012).025**

(.004).015**

(.005)2.036**(.009)

2.963**(.111).132**

(.010).747**

(.067).489*

(.107).481**

(.049)

2.099**(.011).019**

(.004)2.009(.004).061**

(.008)2.282(.091).064**

(.008).476**

(.054).284

(.087).526**

(.040)Marriage formed

Marriage dissolved

No. siblings increases

No. siblings decreases

No. parents

No. children

Sibling within 25 miles

2.004(.012).046*

(.017).010

(.011)2.001(.011)

2.006(.006).006

(.006)

2.007(.173).500**

(.101)2.051(.068).014

(.070)2.002(.059)

2.016(.039)

.011(.067).338**

(.093)2.114(.062)

2.030(.064)

2.028(.052)

2.053(.035)2.750**(.038)

ConstantR2

1.103.111

1.111.111

9.128.100

9.139.100

6.054.352

*p , .01. **p , .001.

and those who are immigrants are less likely tosustain proximity to their siblings, whereas Afri-can Americans and those with more siblings aremore likely to stay proximate. The family lifecourse changes added in Model 1b do not addsignificantly to explained variance, nor does con-trol for these changes substantially alter the effectof age on proximity. As anticipated by the lifecourse model, however, the dissolution of a mar-riage (through widowhood or divorce) is associ-ated with significantly greater proximity. Figure 1depicts the predicted change in likelihood of hav-ing a sibling within 25 miles based on the analysisin Model 1b.

Sibling contact. Consistent with cross-sectionalempirical work, the effect of aging on sibling con-tact is negative. A positive squared term and anegative cubed term are also significant (Model

2a), indicating a flattening of this negative effectin middle age. Sibling contact decreases moreamong the better educated and immigrants andless among those with large sibships, AfricanAmericans, Latinos, and women. Because thesepredictors of contact are very similar to those re-ported in previous cross-sectional research, it ap-pears that these are not spurious correlates andthese individuals are more likely to sustain contactwith siblings.

A comparison of coefficients for the age vari-ables in Models 2a and 2b suggests that familylife course changes account for none of the age-related change in sibling contact—neither thelong-term decline or the modifications in old age.Becoming unmarried between waves, however,predicts significantly greater sibling contact. Thisfinding is consistent with the hierarchical com-pensatory model in which individuals with short-

Page 8: Sibling Relationships Over the Life Course: A Panel Analysis

562 Journal of Marriage and Family

TABLE 2. CHANGING EXCHANGE WITH SIBLINGS: RANDOM EFFECTS MODEL

Receive Help from Siblings Give Help to Siblings

Variable

Sibs Nearby

3a 3b

None Nearby

3c 3d

Sibs Nearby

4a 4b

None Nearby

4c 4d

Age

Age2

(310)Age3

(31000)Education

Immigrant

No. siblings

AfricanAmerican

Latino/Latina

Female

.005(.005)

2.009**(.002).012**

(.002).034**

(.004)2.121(.048).027**

(.004)2.071*(.024).036

(.039).174**

(.019)

.0008(.005)

2.008**(.002).015**

(.002).037**

(.004)2.114(.048).026**

(.004)2.080**(.024).028

(.039).166**

(.019)

2.011**(.002).0009*

(.0003)ns

.026**(.003)

2.036(.030).013**

(.003)2.046(.024)

2.035(.037).117**

(.015)

2.011**(.002).0009*

(.0003)ns

.026**(.003)

2.036(.030).014**

(.003)2.047(.024)

2.035(.037).118**

(.015)

.001(.006)

2.011**(.002).015**

(.002).062**

(.005)2.130(.059).040**

(.004)2.157**(.029)

2.021(.047).134**

(.023)

.007(.006)

2.014**(.002).017**

(.002).062**

(.005)2.126(.059).042**

(.004)2.166*(.029)

2.026(.047).138**

(.023)

2.023**(.002).002**

(.0004)ns

.040**(.004)

2.078(.040).020**

(.004)2.069(.031).015

(.0147).080*

(.020)

2.021**(.002).002**

(.0004)ns

.038**(.004)

2.079(.039).023**

(.004)2.066(.031).014

(.048).086**

(.020)Marriage

formedMarriage

dissolvedNo. siblings

increasesNo. siblings

decreasesNo. parents

No. children

2.048(.037).158**

(.049)2.010(.031).046

(.035)2.109**(.026).077**

(.019)

.023(.029).031

(.042)2.016(.030).051

(.028).038

(.021)2.010(.014)

2.069(.042)

2.041(.057)

2.034(.037).034

(.040)2.107**(.031)

2.035(.022)

2.026(.037)

2.061(.053)

2.034(.037)

2.010(.035).043

(.026)2.071**(.018)

ConstantR2

.166

.074.216.078

.057

.063.016.064

.212

.119.300.121

.231

.095.226.098

*p , .01. **p , .001.

falls in their inner circles bring siblings back clos-er to the center or it could signal simply thatindividuals have time again to spend with siblings.

Change in proximity (added in Model 2c) hasa powerful main effect on sibling contact, andchange in proximity appears to account for ap-proximately one-quarter of the previously ob-served effects of age on sibling contact. Onceproximity is controlled, there is less decline inyoung adulthood, less flattening in middle age,and no significant effect of age cubed (resultsfrom Model 2c are displayed in Figure 1). Withthe effect of proximity removed, the positive re-lationship between education and contact reportedin other studies appears. Proximity explains agood part of the greater contact of African Amer-icans and Latinos and those with marital dissolu-tions. In fact, the greater likelihood that women

will sustain sibling contact is the only effect ap-parently impervious to distance. From these re-sults, we conclude that the greater geographicaldispersion of sibships with age contributes sub-stantially to declines in sibling contact, that lowerlikelihood of dispersion explains a substantial partof the greater sibling contact and exchange amongAfrican Americans, Latinos, and the less well-ed-ucated, and that moving closer to siblings is animportant mechanism through which those expe-riencing marital dissolution sustain greater contactwith their siblings.

Multiplicative interaction terms tested whetherthe age patterns observed here depended on gen-der or on proximity. The gender interactions werenot significant, either individually or as a set, butproximity interactions were significant (F 5 6.1,p , .01). Examination of these effects (not

Page 9: Sibling Relationships Over the Life Course: A Panel Analysis

563Sibling Relationships

FIGURE 1. CONTACT WITH (AND PROXIMITY TO) SIBLINGS BY AGE, 1987–1988 TO 1992–1994

shown) revealed a slight dampening of effectswhen respondent had no sibling within 25 milesbut revealed no change in substance.

Receiving from siblings. Analysis of sibling ex-change relationships is summarized in Table 2.Proximity had a very strong main effect on bothdimensions of exchange, and tests of interactiondemonstrated significantly different age effectsamong close and distant sibships for receiving (F5 24.1, p , .001) and giving (F 5 31.9, p ,.001). Therefore, results for these dependent var-iables are given separately for those with andwithout siblings within 25 miles.

Model 3a in Table 2 reports the results for re-ceiving help from siblings for respondents whohave a nearby sibling. The model demonstrates acurvilinear relationship between age and changesin receiving help from siblings. In proximate sib-ships, receiving help from siblings declines laterand goes down at an accelerating pace during theyears between 20 and 70 and demonstrates a re-bound in old age (Figure 2). Those most apt tosustain support from siblings over time are thosewith more education, larger sibships, and women.African Americans show a pattern of declining aidfrom siblings. Unlike cross-sectional data, whichhave generally shown a lifelong pattern of declin-ing exchange with age, these time series data sug-gest that some behavioral measures of sibling sol-idarity increase in old age.

Three of the life course changes included in

this analysis are associated with changes in sup-port from nearby sibs (Model 3b). Receiving helpfrom sibs is greater for those whose marriages dis-solved, lower for those whose parents remainedalive, and greater for those who add children.Controlling for these life course changes, how-ever, does not account for the observed age ef-fects.

The pattern for those with no nearby siblingsis somewhat different (see Figure 2). The down-ward pattern during early and middle adulthood isdampened, and there is no upturn in old age(Model 3c). None of the family life course chang-es are related to changes in receiving support inthese dispersed sibships, and only gender, sibshipsize, and education are related significantly to sus-tained support from siblings (Model 3d). Eventhese robust effects, however, are weaker for dis-persed than for proximate sibships.

Giving support to siblings. The story for giving tosiblings is almost identical to that for receiving.Giving to siblings declines in early and middleadulthood but rebounds notably in old age amongrespondents who have at least one nearby sib(Model 4a and Figure 2). Changed giving to sib-lings is associated with many of the same factorsthat predicted changed receipt of support. Thosewho have higher educations, more siblings, or arefemale are more likely to sustain or increase giv-ing to siblings between waves, whereas AfricanAmericans are less likely to sustain giving. As be-

Page 10: Sibling Relationships Over the Life Course: A Panel Analysis

564 Journal of Marriage and Family

FIGURE 2. GIVING AND RECEIVING SIBLING SUPPORT BY AGE AND PROXIMITY 1987–1988 TO 1992–1994

fore, these effects are weaker among those whohave no close siblings.

Life course changes added in Models 4b and4d do not explain the age effects and have fewmain effects on giving to siblings. Among thosewith nearby siblings, those whose parents remainalive give less to siblings. Among those withoutnearby siblings, increased number of children isassociated with less giving to siblings. Althoughthe positive effect of larger numbers of childrenon receipt of support in Model 3b could have in-dicated either that those with more children hadgreater family involvement or greater need, thenegative effect for giving suggests that number ofchildren triggers support from one’s siblings andsimultaneously allows one to justify less giving.

Fixed Effects Replication

The random effects solution used in the foregoinganalyses provides efficient estimation of coeffi-cients if it is assumed that the model is completelyspecified and we have omitted no critical time-invariant variables. In this case, it seems plausiblethat unmeasured and relatively invariant attitudes,such as feelings of family obligation or family af-

fection, could bias the results. To test this possi-bility, I replicated the analysis using a fixed effectssolution. The fixed effects solution looks only atwithin-individual change and effectively controlsfor all variables—measured and unmeasured—that remain fixed within individuals, including notonly attitudes and values but also cohort, gender,race, and ethnicity. The fixed effects models,therefore, included only age and the life coursemeasures.

The fixed effects models, summarized in Table3, show age effects that are virtually identical tothose presented using a random effects solution.All four measures are characterized by long-termdecline with slight rises in old age for contact,receiving, and giving. The strong similarity of re-sults supports the conclusion that the age effectsdiagramed in Figures 1 and 2 cannot be accountedfor by unobserved differences between individu-als.

Although the general implications of the ran-dom- and fixed-effects solutions are very similar,a few of the life course changes differ betweenthe two solutions. Two life course changes are sig-nificant in the fixed effects solution that were in-significant in the earlier analysis. Marital forma-

Page 11: Sibling Relationships Over the Life Course: A Panel Analysis

565Sibling Relationships

TABLE 3. CHANGING SIBLING SOLIDARITY BETWEEN WAVES: FIXED EFFECTS SOLUTION

Variable Proximity Contact Receiving Giving

Age

Age2 (310)

Age3 (31000)

Marriage formed

Marriage dissolved

No. siblings increased

No. siblings decreased

No. parents

No. children

Proximity

2.016**(.003).005**

(.001)2.005*(.001)

2.016(.014).049*

(.019).077**

(.012)2.050**(.013)

2.008(.013).019

(.009)

2.101**(.017).032**

(.006)2.026**(.007)

2.242*(.084).436**

(.113).127

(.072)2.086(.079)

2.109(.081).113

(.055)1.924**(.064)

.002(.007)

2.006(.003).008*

(.003)2.087*(.034).114

(.045).077*

(.029).065

(.032)2.057(.032).045

(.022).210**

(.025)

2.0001(.008)

2.010**(.003).012**

(.003)2.059(.039)

2.084(.052).073

(.034).034

(.037)2.102*(.037)

2.057(.025).263**

(.029)

ConstantOverall R2

.716

.0127.3752.73

.494

.0691.040.109

*p , .01. **p , .001.

tion is associated with reduced contact andreducing receiving from siblings in the fixed ef-fects analysis, and reporting an increased numberof siblings is associated with increased receivingfrom siblings. Three variables that were signifi-cant in the random effects solution do not reachsignificance in the fixed effects solution: maritaldissolution and retaining parents have smaller ef-fects on changes in receiving support and in-creased number of children is less strongly relatedto reduced giving support. The finding that in-creased number of children sustains proximity tosibs may suggest a deliberate choice to stay closeto family during the early stages of family build-ing, but overall the findings support the conclu-sion that life courses changes have occasionallysignificant, but small and inconsistent effects onbehavioral measures of sibling relationships thatare generally in the direction suggested by lifecourse models.

DISCUSSION

This analysis uses a large national panel data setto estimate the effects of aging and family lifecourse changes on four measures of sibling rela-tionships—proximity, contact, giving help, and re-ceiving help—measured approximately 5 years

apart. The dominant pattern of sibling relation-ships over the adult life span is reduced contactand exchange coupled with strong staying power.Proximity and contact decrease modestly in earlyadulthood and then show long-term stabilitythrough old age. Giving and receiving help de-cline from age 16 until old age, however a sub-stantial resurgence in sibling exchange is dem-onstrated after age 70 among those with nearbysiblings. These age patterns are found in both ran-dom and fixed effects pooled time series solutions,suggesting that they are quite robust and that ob-served patterns are not likely to be accounted forby unmeasured invariant characteristics includingcohort effects. These age effects are also largelyimpervious to controls for changes in proximityand in the family life course.

Family life course effects are smaller and lessconsistent than age effects but significant effectsare in the expected directions: sibling contact andexchange are increased by marital dissolution anddecreased by additional children, new marriages,retention of parents, and growing distance. Thedirection of life-course effects is generally consis-tent with the notion that siblings are members ofthe second tier in adults’ family networks, whoseprominence depends to a significant extent on theinner circle. With the exception of contact, how-

Page 12: Sibling Relationships Over the Life Course: A Panel Analysis

566 Journal of Marriage and Family

ever, these life course variables account for littleor none of the observed age effects, and it doesnot appear to be true that widowhood and deathof parents play important roles in increased siblingcontact in old age or that marriage and growingnumbers of children explain the decline in siblingexchange and contact in the adult years.

Several findings in this analysis raise questionsabout the generality of the hierarchical model inwhich siblings are pushed to the second tier byevents during adult life. Contrary to life coursepredictions, adding children is more likely to en-ergize than diminish sibling relationships, a find-ing consistent with Connidis’s (1992) work on anover-55 population. This suggests that becominga parent encourages people to maintain ties withtheir siblings, perhaps so that their children haveactive relationships with aunts, uncles, and cous-ins. In addition, several categories of individualsseem inclined to maintain and strengthen their sib-ling relationships during adulthood regardless ofevents in their own lives. Generalizing across alldependent variables, these individuals are women,those who continue to live close to their familiesof origin, who have large sibships, and higher ed-ucations. These findings support several patternsfound repeatedly in the family literature, the gen-der-based nature of kinkeeping, the importance ofproximity, and the positive effect of socioeconom-ic status on kin exchange. Taken together, thesefindings support the argument that a simple hier-archy of obligations is insufficient to explain pat-terns of sibling relationships. For some tasks andfor some individuals, a brother or sister might bepreferred to a spouse or an adult child, and rela-tionships with siblings are sustained despite ageand changing circumstances.

The effects of aging itself are consistent andprovide broad support for a developmental per-spective. The declines in sibling proximity andcontact in early adulthood are consistent with thehypothesis that a reduced sibling bond is part ofa nearly universal individuation process in whichindividuals increase the physical and psychologi-cal distance between themselves and their familiesof origin. Despite this disengagement process, sib-lings remain life long parts of most adults’ socialnetworks. On the average, they keep track of oneanother by talking or visiting once or twice amonth, and levels of contact are remarkably stablefrom age 45 to 85. Although relatively small lev-els of goods and services pass between adult sib-lings, sibling exchange picks up significantly inold age among those with nearby siblings. This

finding supports the suggestion of developmentaltheory that siblings appear to take on greater im-portance as elders search for contemporaries withwhom to share old stories and reaffirm the past.As society becomes more mobile, siblings maytake on increased importance. Where few remainin the communities in which they were born—andif they have, their neighbors haven’t—siblingsmay be the only members of older individuals’social networks who can not only rememberWorld War II, but who can also remember themin 1940.

The conceptual literature generally has beencareful to stipulate that the upsurge in sibling re-lationships in old age is likely to be confined tomeasures of attachment and affection and to de-velop despite reduced distance, contact, and actualexchange (Goetting, 1986). The present analysis,however, demonstrates modest upturns in mea-sures of giving support and receiving support.Given the consistency with which cross-sectionalresearch has shown negative associations with agein these behavioral measures, even the modest up-turns found in old age in this analysis are note-worthy. Presumably, if measures of sibling attach-ment were available, these measures would showeven stronger increases.

How important can these upturns in sibling ex-change and contact be in the lives of the averageelder? Sibling numbers suffer substantial attritionin old age (nearly 50% of the respondents over 70reported fewer siblings at Time 2 than at Time 1),and nearly half of those in this age group no lon-ger have a sibling within 25 miles. Despite theseobstacles, there is a strong pattern of contactmaintenance and a consistent pattern of elders ex-changing more with nearby siblings in old age. Itis known that, at all ages, well-being is enhancedmore by feelings of closeness and attachment thanby direct assistance (House, Umberson, & Landis,1988), and it is possible that even modest rela-tionships with siblings can make important con-tributions to morale. The durable sibling relation-ship is ideally suited to support life review and toprovide conversation about a past that may seemmore interesting than the present (Seltzer, 1989).Especially for those without partners, a sibling re-lationship may thwart a disconcerting sense thatthey have outlived their generation.

Siblings are not major actors in the averageadult’s day-to-day life. If one takes a long-rangeview, however, and looks back at those who haveplayed a part in one’s life for 40, 60, or 80 years,the sibling relationship takes on a more central

Page 13: Sibling Relationships Over the Life Course: A Panel Analysis

567Sibling Relationships

role. It is a relationship that most people maintainand nurture—or in some cases, perhaps, just en-dure—throughout the entire life span and which,in old age, gains increasing prominence. The rel-ative weakness of life course effects suggests thatthe latter is not simply because siblings are a sec-ond-rate replacement for losses among preferredassociates. Instead, it appears that, despite grow-ing distance and loss, siblings provide somethingthat more recent members of the family circle can-not.

NOTE

Naomi Lacy, Hongyu Wang, and Kurt Johnson providedmuch appreciated technical support for this project. Iam indebted to David R. Johnson and Agnes Riedmannfor insightful suggestions on preliminary drafts and tothose from anonymous reviewers. This article was pre-sented at the 1997 meetings of the American Sociolog-ical Association in Toronto, Ontario, Canada.

REFERENCES

Allan, G. (1977). Sibling solidarity. Journal of Mar-riage and the Family, 39, 177–184.

Allison, P. (1994). Using panel data to estimate the ef-fect of events. Sociological Methods and Research,23, 174–199.

Antonucci, T., & Akiyama, H. (1987). Social networksin adult life and a preliminary examination of theconvoy model. Journal of Gerontology, 42, 519–527.

Avioli, P. (1989). The social support functions of sib-lings in later life: A theoretical model. American Be-havioral Scientist, 33, 45–57.

Bengtson, V. L., & Roberts, R. E. L. (1991). Intergen-erational solidarity in aging families: An example offormal theory construction. Journal of Marriage andthe Family, 53, 856–870.

Bowen, M. (1976). Theory and practice n psychother-apy. In P. J. Guerin (Ed.), Family therapy: Theory andpractice. New York: Gardner Press.

Cantor, M. (1979). Neighbors and friends: An over-looked resource in the informal support system. Re-search on Aging, 1, 434–463.

Cicirelli, V. G. (1985). Sibling relationships throughoutthe life cycle. In L. L’Abate (Ed.), The handbook offamily psychology and therapy (Vol. 1, pp. 177–214).Homewood, IL.: Dorsey Press.

Cicerelli, V. G. (1996). Sibling relationships in middleand old age. In G. Brody (Ed.), Sibling relationships:Their causes and consequences. Advances in AppliedDevelopmental Psychology, (Vol. 10, pp. 47–74).Westport, CN: Ablex.

Connidis, I. A. (1989). Siblings as friends in later life.American Behavioral Scientist, 33, 81–93.

Connidis, I. A. (1992). Life transitions and the adultsibling tie: A qualitative study. Journal of Marriageand the Family, 54, 972–982.

Connidis, I. A. (1994). Sibling support in older age.Journal of Gerontology, 49, S309–S317.

Connidis, I. A., & Campbell, L. D. (1995). Closeness,

confiding and contact among siblings in middle andlate adulthood. Journal of Family Issues, 6, 722–747.

Connidis, I. A., & Davies, L. (1990). Confidants andcompanions in later life: The place of family andfriends. Journal of Gerontology, 45, S141–S149.

Connidis, I. A., & Davies, L. (1992). Confidants andcompanions: Choices in later life. Journal of Ger-ontology, 47(3), S115–S122.

Dunn, J. (1984). Sibling development and the devel-opmental impact of critical incidents. In P. B. Baltes& O. G. Brim, Jr. (Eds.), Life-span development andbehavior (Vol. 6, pp. 335–353). New York: AcademicPress.

Elder, G. H., Jr. (1974). Children of the great depres-sion. Chicago: University of Chicago Press.

Erikson, E. H. (1959). The problem of ego identity. Psy-chological Issues, 1, 101–164.

Fischer, C. S. (1982). To dwell among friends: Personalnetworks in town and city. Chicago: University ofChicago Press.

Goetting, A. (1986). The developmental tasks of si-blingship over the life cycle. Journal of Marriage andthe Family, 48, 703–714.

Gold, D. T. (1989). Sibling relationships in old age: Atypology. International Journal of Aging and HumanDevelopment, 28, 37–51.

Hagestad, G. O. (1986). The family: Women and grand-parents as kin-keepers. In: A. Piper & L. Bronte(Eds.), Our aging society (pp. 141–160). New York:Norton.

Harvey, D. M., & Bray, J. H. (1991). Evaluation of anintergenerational theory of personal development:Family process determinants of psychological andhealth distress. Journal of Family Psychology, 4,298–325.

Hetherington, E. M. (1994). Siblings, family relation-ships, and child development: Introduction. Journalof Family Psychology, 8, 251–253.

Hohn, C., & Mackensen, R. (1989). Introduction. In E.Grebenik, C. Hohn, & R. Mackensen (Eds.), Laterphases of the family life cycle: Demographic aspects(pp. 1–18). Oxford, U.K.: Clarendon.

House, J. C., Umberson, D., & Landis, K. R. (1988).Structures and processes of social support. AnnualReview of Sociology, 14, 293–318.

Johnson, C., & Catalano, D. (1981). Childless elderlyand their family supports. Gerontologist, 21, 610–618.

Lee, E. S., Forthofer, R. N., & Lorimer, R. J. (1989).Analyzing complex survey data. Newbury Park: Sage.

Lee, T. R., Mancini, J. A., & Maxwell, J. W. (1990).Sibling relationships in adulthood: Contact patternsand motivations. Journal of Marriage and the Fam-ily, 52, 431–440.

Lopata, H. Z. (1973). Widowhood in an American city.Cambridge, MA: Sheckman.

Miner, S., & Uhlenberg, P. (1997). Intragenerationalproximity and the social role of sibling neighbors af-ter midlife. Family Relations, 46, 145–153.

Parsons, T. (1943). The kinship system of the contem-porary United States. American Anthropologist, 45,22–38.

Reiss, D., & Oliveri, M. E. (1983). The family’s con-struction of social reality and its ties to its kin net-work. Journal of Marriage and the Family, 45, 81–91.

Page 14: Sibling Relationships Over the Life Course: A Panel Analysis

568 Journal of Marriage and Family

Riedmann, A., & White, L. K. (1996). Racial and ethnicdifferences in sibling relations. In G. Brody (Ed.),Sibling relationships: Their causes and consequences.Advances in Applied Developmental Psychology, 10,105–126.

Rosenberg, G. S., & Anspach, D. F. (1973). Sibling sol-idarity in the working class. Journal of Marriage andthe Family, 35, 108–113.

Rossi, A., & Rossi, P. (1990). Of human bonding. NewYork: Aldine de Gruyter.

Scott, J. P. (1983). Siblings and other kin. In T. Brubaker(Ed.), Family relationships in later life (pp. 47–62).Beverly Hills, CA: Sage.

Seltzer, M. (1989). The three R’s of life cycle sibships:Rivalries, reconstuctions, and relationships. AmericanBehavioral Scientist, 33, 107–115.

Simons, R. L. (1983–1984). Specificity and substitutionin the social networks of the elderly. InternationalJournal of Aging and Human Development, 18, 121–139.

Suggs, P. (1985). Discriminators of mutual helping be-havior among older adults and their siblings. Journalof Applied Gerontology, 4, 63–70.

Sweet, J., Bumpass, L., & Call, V. (1988,). The designand content of the national survey of families andhouseholds (Working Paper No. NSFH-1). Madison:University of Wisconsin Center for Demography andEcology.

Taylor, R. J., Chatters, L., & Mays, V. (1988). Parents,children, siblings, in-laws, and non-kin as sources ofemergency assistance to Black Americans. FamilyRelations, 37, 298–304.

Treas, J., & Bengston, V. (1987). The family in lateryears. In M. Sussman and S. Steinmetz (Eds.), Hand-book of marriage and the family (pp. 625–648). NewYork: Plenum.

Wellman, B., & Wortley, S. (1990). Different strokesfrom different folks: Community ties and social sup-port. American Journal of Sociology, 96, 558–588.

White, L. K. (1998). Who’s counting? Quasi-facts andstepfamilies in reports of number of siblings. Journalof Marriage and the Family, 60, 725–733.

White, L.K., & Riedmann, A. (1992). Ties among adultsiblings. Social Forces, 71, 85–102.

Winship, C., & Radbill, L. (1994). Sampling weightsand regression analysis. Sociological Methods andResearch, 23, 230–245.