Copyright 1998 by
The Cerontological Society of America
The Cerontologist
Vol. 38, No. 3, 276-285
Using data from the second wave of the National Survey of Families and Households, the
authors examine the determinants of frequent and infrequent visiting between grandparents and their grandchild sets. A grandchild set consists of all the children of
a particular child of the grandparent, provided that the grandchildren reside in
their parent's household. The 6 significant predictors of frequent or infrequent
contact are geographic distance, quality of relationship between grandparent and
parent of the grandchild set, number of grandchild sets, gender of grandparent,
lineage of the grandchild set, and marital status of the grandparent.
Key Words: Kinship, Lineage, Intergenerational relationships
Frequency of Grandparent Contact
With Grandchild Sets: Six Factors
That Make a Difference
Peter Uhlenberg, PhD,1 and Bradley G. Hammill :
The potential for grandparents to have frequent,
face-to-face contact with their young grandchildren
is probably greater today than at any previous time
in history (Uhlenberg & Kirby, in press). Moreover,
national studies of intergenerational relationships
repeatedly find that most grandparents do have frequent contact with at least some of their grandchildren (Cherlin & Furstenberg, 1986; Gravenish &
Thomson, 1996; Harris & Associates, 1975). Nevertheless, these studies also reveal substantial variability among grandparents in level of involvement
with their grandchildren. At the extremes, some see
grandchildren daily whereas others have no contact.
This article uses data from a large, national survey to
examine several critical variables predicting whether
grandparents have either frequent or infrequent contact with various sets of their grandchildren.
A new study is not needed to establish that geographic distance is a strong predictor of frequency of
intergenerational contact. Every study that includes
a measure of distance finds that the closer grandparents live to grandchildren, the more likely they are to
visit frequently (Cherlin & Furstenberg, 1986; Hodgson,
1992; Whitbeck, Hoyt & Huck, 1993). Further, these
same studies have established that the quality of the
relationship between grandparents and their children (G1-G2) is related to the frequency of contact
between grandparents and their grandchildren ( G I GS). Past research has not, however, provided clear
and/or consistent findings regarding the role of a
Address correspondence to Peter Uhlenberg, PhD, Department of
Sociology, CB #3210, University of North Carolina, Chapel Hill, Chapel
Hill, NC 27599-3210. E-mail: [email protected]
department of Sociology, University of North Carolina, Chapel Hill.
276
number of other potentially interesting variables. The
lack of clear findings is because most studies are based
on small and/or nonrepresentative samples (Aldous,
1995; Robertson, 1995). Using a nationally representative sample of more than 4,600 grandparent-grandchild set dyads, we are able to more carefully assess
the impact of factors that potentially make a difference in frequency of contact.
Background
On the basis of previous theoretical and empirical
studies of intergenerational relationships, there are a
number of variables that might be expected to influence the frequency of contact between grandparents
and grandchildren. We will group these variables in
several categories: opportunity variables; gender/family
variables; and sociodemographic variables.
Opportunity Variables.—Geographic distance is
the most obvious factor influencing opportunity for frequency contact. Both the time and expense of maintaining frequent face-to-face contact increase as distance between grandparents and grandchildren
increases. As already noted, studies routinely find distance to be by far the strongest of all predictors of
contact.
A second variable affecting opportunity for frequent interaction is the number of grandchildren an
older person has. Given that time and resources are
limited, one might expect that as the number of
grandchildren increases, the opportunity for grandparents to have frequent contact with any particular
grandchild would decrease. Although this is a plausible
hypothesis, little empirical attention has been given
to the effects of number of grandchildren. In this study,
The Gerontologist
we focus not on individual grandchildren but on
grandchild sets. A grandchild set is defined as the
group of children (G3) living in the household of a
grandparent's (G1) child (C2). Thus, the number of
grandchild sets will vary depending on how many
children (C2) with children (G3) in their households a
grandparent has. Each grandparent-grandchild set
dyad is examined separately, and the effect of number of grandchild sets on frequency of visiting with a
particular grandchild set is determined. This focus on
grandchild sets provides a useful approach to studying involvement of grandparents with their grandchildren, although it cannot be used to explore some
interesting facets of the relationship. For example,
using this approach, the role of such individual grandchild characteristics as age and gender cannot be
examined directly.
Two additional variables may affect opportunity
for frequent contact, although previous research has
not found them to be significant (Cherlin & Furstenberg, 1986). Grandparents' health status could plausibly be a factor if poor health limits mobility. However, because the most impaired grandparents, those
living in nursing homes, are excluded from our sample,
this is not expected to be an important variable. The
other variable, labor force status of the grandparent,
also could have an effect on frequency of contact,
assuming that those in the labor force have less time
available for visiting.
Gender/Family Variables.—No discussion of grandparenthood in America should omit consideration
of the role of gender. The importance of gender in
kinship relationships was stressed by Alice Rossi
when she wrote, "As childrearers, caregivers, and kin
keepers, women provide the glue that holds family
and lineages together" (1995, p. 275).
Hagestad (1986) made a similar point when she
discussed the kin-keeping role of women and the increasing "female axis" of intergenerational relationships. In considering how gender might affect
frequency of contact between grandparents and sets
of grandchildren, both the grandparents' sex and the
lineage of the grandchildren may be important.
Given the kin-keeping role of grandmothers, we expect they will have more frequent contact with
grandchildren than do grandfathers. And if the female
axis of intergenerational linkages is most important,
we would expect greater contact between maternal
grandparents and grandchildren than between paternal ones. Several studies have supported this hypothesis (Gravenish & Thomson, 1996; Roberto & Stroes,
1992; Rossi & Rossi, 1990). It is surprising, however,
that several other studies report that frequency of
contact was not associated with gender of the grandparent and that there was no significant difference
between maternal and paternal grandparents (Cherlin
& Furstenberg, 1986; Hodgson, 1992).
It is expected that marital status will be a significant
predictor of frequency of contact with grandchildren
for grandfathers. It is clear that divorced older men
have far less contact with their adult children than
married ones (Cooney & Uhlenberg, 1990). Further,
Vol. 38, No. 3,1998
277
widowed grandfathers may have less contact with
their grandchildren than married ones because they
lack a wife to facilitate the maintenance of close kinship ties. Marital status of grandmothers also may
be significant in some cases, either because lack of a
spouse may weaken ties to children or reduce resources needed for traveling to a child's house. We
anticipate, however, that marital status will have
greater consequences for contact with grandchildren
among men than women.
One additional aspect of kinship relationships is
considered in this study: the quality of the relationship between the parent and the grandparent of
the grandchild set. The middle generation has been
characterized as the "bridge" or "gatekeeper" for
the G1-G3 relationship, with the ability to facilitate
or impede the frequency of interactions. When parents have an emotionally close relationship with the
grandparents it is reasonable to expect that the
grandparents will see that set of grandchildren more
frequently than when the parent-grandparent relationship is strained. Several studies, in addition to
those cited above, provide support for the hypothesis
that positive affect between the first and second generations promotes closeness and contact between
the first and third generations (Johnson, 1985; King &
Wider, 1995; Matthews & Sprey, 1985).
Sociodemographic Variables.—Several standard
control variables are included in the cluster of sociodemographic factors, although no firm hypotheses
regarding their effects are suggested by previous research. The three variables included in this cluster are
grandparents' age, race, and education. Although age
is not expected to have much effect, it is possible
that older grandparents will tend to have less stamina
than younger ones (even after controlling for health
status), and hence will be less likely to visit frequently
with their children and grandchildren. Possible racial
differences in kinship ties have been of interest to
researchers, and it is clear that African American
grandparents are more likely than Whites to function
as surrogate parents (Hernandez, 1993; Szinovacz,
1998). Among grandparents who do not coreside
with grandchildren, however there is little empirical
evidence suggesting that African Americans are
more likely to have frequent contact with grandchildren (Cherlin & Furstenberg, 1986). Finally, possible
effects of educational attainment are examined. Some
suggest that with increasing education there is a
tendency for individuals to rely less on kin for social
relationships (Fischer, 1982). On the other hand,
higher socioeconomic status may be associated with
more resources, which could facilitate more frequent
visiting.
On the basis of results of previous empirical studies
and/or theoretical discussions of kinship and gender,
we expect to find a number of significant predictors
of frequent or infrequent contact between grandparents and their sets of grandchildren:
1. Geographic proximity is expected to be a strong
predictor of frequency of face-to-face contact.
2. Number of grandchild sets is expected to be
negatively associated with frequent contact.
3. Grandmothers are expected to have more frequent contact than grandfathers with grandchild sets.
4. Maternal grandparents are expected to have
more frequent contact than paternal grandparents. Further, the effect of lineage is expected to be greater
for grandmothers than grandfathers, given the saliency
of the mother-daughter link.
5. Higher emotional affect between grandparents
and their children is expected to be a predictor of
more frequent contact between grandparents and grandchildren.
6. Married grandparents are expected to have
more frequent contact with grandchildren than widowed or divorced ones. The effect of marital status is
expected to be larger for grandfathers than grandmothers.
Data and Methods
The data analyzed in this study are from the second wave of the National Survey of Families and
Households (NSFH2). This 1992-93 survey involved
interviews with a national sample of 10,008 adults.
From these adults, we selected all grandparents (G1)
who had at least one child (G2) who had one or more
children (G3) living in his/her household. We are interested not in individual grandchildren, but in each
set of grandchildren that a grandparent has. A set
includes all of the grandchildren living with a particular adult child of the grandparent. Thus, a grandparent will have as many sets of grandchildren as she/
he has adult children who have children living at
home. Excluding those who coreside, the NSFH2 contains complete information for 4,629 grandparentgrandchild set dyads.
The outcome variable of interest in this study is frequency of face-to-face contact between a grandparent and a particular set of grandchildren. The
measure of frequency of contact comes from a question asking about frequency of contact between the
grandparent and the parent of the grandchild set (who
is the son or daughter of the grandparent). Although
this is not an ideal measure, it provides a reasonable
indicator of how frequently grandparents have contact with a particular set of their grandchildren. Grandparents in NSFH2 were asked how frequently they
had contact with grandchildren (on a 6-point scale),
as well as how frequently they had contact with each
child (using the same scale). In most cases where the
grandparent had just one child who had children in
his/her household, the contact reported with grandchildren provides a direct measure of contact with the
grandchild set of the child with children. (This is not
true in all cases, however. See Appendix.) The correlation between reported contact with the child and
with the grandchild set is 0.72, and in 86% of the
cases the reported frequency of contact with the child
and the grandchildren is either identical or in an
adjacent category. (In cases where there is greater
discrepancy, 5% of the grandparents report more
frequent contact with their child; 10% report more
278
frequent contact with their grandchildren.) For grandparents with one set of grandchildren, we compared
multivariate analyses where the dependent variable
(contact with grandchild set) is measured in these two
ways. The results are quite similar (see Appendix).
Two dependent variables are constructed from the
6-point scale of frequency of visiting with the parent
of each grandchild set: 1 = no visiting in the last year,
2 = once a year, 3 = several times a year, 4 = 1 to
3 times a month, 5 = once a week, 6 = more than
once a week. "Frequent" contact is defined as visiting weekly or more often (5 or 6 on the scale); "infrequent" contact is defined as visiting once a year or
less (1 or 2 on the scale). The decision to use these
two dependent variables developed after we ran
an ordered logistic regression and found that the
proportional odds assumption was violated, x2 (52,
N = 4608) = 371, p < .001. This indicates that some
independent variables are operating differently at
different levels of the dependent variable. We rejected the alternative of using multinomial logistic
regression because results for such an analysis are
inordinately complicated to discuss with a six-level
variable. By creating two variables—one for each
end of the distribution—we are able to discuss results
clearly and simply. Also, by using simple logistic regression we are able to use all of the data, correcting
for family clustering by using Huber standard errors.
As indicated in the preceding section, a number of
relatively straightforward independent variables are
included in this study. Grandparents' age, race (White,
Black, Other), education (years of schooling completed), work status (in the labor force, not in the
labor force), and health (5-point scale, from "excellent" to "very poor") are included primarily as control
variables, with no strong expectations regarding their
effects. Grandparents' gender (female, male), kinship link to the grandchild set (maternal, paternal),
quality of relationship with the G2 parent (10-point
scale, from "very poor" to "excellent"), and marital
status are critical variables in the family/gender category. The four categories of marital status used in
the multivariate analysis are married (meaning married to the other parent of the son or daughter who
is parent of the grandchild set), remarried (meaning
current spouse is not the parent of the relevant child),
widowed, and divorce/separated (which also includes
the 37 grandparents who were never married). Distance is measured in miles between the grandparent
and grandchild set, and log of distance is used in the
regression analysis. Finally, number of grandchild sets
(1, 2, 3, 4+) is included as a predictor variable. In a
preliminary analysis we also included several characteristics of G2 (marital status, labor force participation)
as independent variables. Because none of these was
a significant predictor, and to keep the focus on
grandparents, we dropped these variables in subsequent analysis. The distribution of the sample by the
various independent variables is shown in Table 1.
We examine logistic regression equations for two
different dichotomous outcome variables: frequent
versus not-frequent contact and infrequent versus
not-infrequent contact. We initially ran these regres-
The Gerontologist
Table 1. Distribution of Grandparents of Grandchild Sets
on Selected Variables and Percent of Grandparents
With Various Characteristics Who Have Frequent
or Infrequent Contact3 With Grandchild Sets
Sample
Distribution
Grandparent
Characteristics
Percent With
%
n
Frequent
Contact0
Infrequent
Contact
24.6
33.4
42.0
1138
1546
1945
52.4
42.2
40.8
12.8
16.8
14.9
Gender
Grandmother
Grandfather
66.3
33.7
3069
1560
47.4
37.8
14.0
17.1
Kin Link
Maternal
Paternal
58.4
41.6
2703
1926
47.3
39.7
13.1
17.8
Race
White
Black
Other
73.9
18.1
7.9
3423
480
366
42.1
49.2
51.4
13.8
19.9
15.3
Education
12 years or less
13 years or more
75.6
24.4
3499
1130
46.3
37.4
15.5
13.6
Work status
In labor force
38.5
1780
2849
46.2
42.9
13.9
15.8
1897
739
1109
884
49.4
38.0
42.3
39.8
9.2
19.4
17.0
22.1
Age
<55
55-64
65 +
Not in labor force
61.5
Married
41.0
Remarried
16.0
24.0
19.1
Divorced/separated
Health
Very poor to poor
Fair to excellent
9.9
90.1
457
4172
44.0
45.7
14.8
16.9
Relationship with child
Very poor to fair
Good to excellent
7.7
92.3
356
4273
32.0
45.1
30.1
13.8
1810
1074
877
79.3
51.8
4.6
14.0
1.8
3.1
14.8
57.4
1098
1365
972
1194
49.7
43.6
43.6
40.0
12.5
15.1
13.5
18.6
Distance to grandchild set
<10 miles
10-49 miles
50-499 miles
500+ miles
39.1
23.2
18.8
18.9
Number of grandchild sets
1
23.7
2
29.5
21.0
3
25.8
4+
868
Note: Percentages may not sum to 100 because of rounding.
Frequent means once per week or more; infrequent means
once per year or less.
b
Married means living with spouse who is also the parent of
the relevant middle-generation child; remarried means spouse is
not the parent of the relevant child; divorced/separated includes
37 grandparents who reported being never married.
c
Each independent variable, when included in bivariate logistic
regression with frequent contact, is statistically significant at the
.05 level except work status, widowhood, and health.
d
Each independent variable, when included in a bivariate logistic regression with infrequent contact, is statistically significant at
the .05 level except age, work status, widowhood, and "other"
race.
a
Vol. 38, No. 3,1998
Findings
Cross-Tabulation Analysis
Marital status6
Widowed
sions on the entire sample (including both grandmothers
and grandfathers). However, a likelihood ratio test between the original model and a fully interactive model
by grandparents' sex confirmed our expectation that
the analysis should be done separately by gender. A
similar test by race did not indicate that different
processes were operating for Whites, Blacks, or other
races. Therefore, we present four regression models
predicting frequent contact for grandmothers, frequent contact for grandfathers, infrequent contact
for grandmothers, and infrequent contact for grandfathers. Because a number of grandparents are included in more than one grandparent-grandchild set
dyad, we correct for family clustering by using Huber
standard errors. With this correction it is possible to
use each eligible dyad in the regression analysis.
Before considering results from a multivariate
analysis, it is useful to note several interesting findings
from simple cross tabulations. First, as expected, the
more sets of grandchildren that a grandparent has,
the less likely the grandparent is to have frequent
contact with any particular set of grandchildren and
the more likely he/she is to have infrequent contact
with any particular set (see Table 1). Several additional
relationships between number of sets and frequency
of contact are shown in Figures 1 and 2. (Note that
the unit of analysis in Figures 1 and 2 is the individual
grandparent, n = 2,723. This differs from Tables 1
and 2 and Figure 3, where the unit of analysis is the
grandparent-grandchild set dyad, n = 4,629.) As the
number of sets increases, the likelihood that a grandparent has frequent contact with at least one set goes
up. For example, the proportion reporting frequent
contact with at least one set of grandchildren increases from 50% of grandparents with one set to 78%
of those with four or more sets. Similarly, the frequency
of having infrequent contact with all sets of grandchildren declines from 13% of those with one set to only
2% of those with four or more sets. Clearly grandparents who have multiple sets of grandchildren
nave much greater total contact with grandchildren
than do grandparents with only one set. On the other
hand, having multiple sets of grandchildren increases
the chances that some sets will be seen on an infrequent basis. When a grandparent has four or more
sets of grandchildren, it is more than three times as
likely that there is infrequent contact with at least
one set than if the grandparent has only one set
(47% vs 13%). Also, the probability of frequent contact with all sets drops from 50% when there is one
set to only 6% when there are four or more. From the
perspective of an older person, having a large
number of grandchild sets implies a high level of total
interaction with grandchildren. From the perspective
of grandchildren, however, having multiple sets of
cousins reduces the likelihood of frequent contact
with grandparents. Although these findings are not
unexpected, they point out a serious limitation of
279
AT LEAST ONE SET
ALL SETS
90
ANY PARTICULAR SET
T
80-
70-
60-
50-
40-
30-
20-
10-
4+
1
2
3
4+
4+
Number of Sets
Figure 1. Variations in percent of grandparents visiting sets of grandchildren frequently by number of grandchild sets. The difference
between having one set and four sets is statistically significant at the .05 level in all comparisons.
50
AT LEAST ONE SET
ALL SETS
ANY PARTICULAR SET
T
4 5 ••
4+
1
2
3
Number of Sets
Figure 2. Variations in percent of grandparents visiting sets of grandchildren infrequently by number of grandchild sets. The difference between having one set and four sets is statistically significant at the .05 level in all comparisons.
280
The Gerontologist
60 x
Maternal GM
FREQUENT CONTACT
Paternal GM
Maternal GF
INFREQUENT CONTACT
Paternal GF
Maternal GM
Paternal GM
Maternal GF
Paternal GF
Gander and Lineage
Figure 3. Variations in percent of grandparents visiting a particular set of grandchildren frequently or infrequently by gender and
lineage of grandparent. Comparisons between maternal grandmother (GM)/paternal grandfather (CF) and paternal CM/maternal CF are
not particularly interesting. For the remaining eight comparisons, differences are statistically significant at .05, except between maternal/
paternal grandfather for frequent contact.
studies based on samples that include only the relationship between a grandparent and a focal child.
A second noteworthy finding from cross-tabulation
analysis concerns differences by gender and lineage.
As discussed above, previous studies have produced
inconsistent findings regarding differences between
grandmothers and grandfathers and differences between maternal and paternal grandparents. Results
from analyzing this large, national sample indicate that
both the grandparent's gender and lineage do matter, although not all of the differences are Targe (Table
1, Figure 3). The grandparent most likely to have
frequent contact with any particular set of grandchildren is, as expected, the maternal grandmother. About
52% of maternal grandmothers have frequent contact
with any particular set of grandchildren, compared to
41% of maternal grandfathers. Further, maternal grandmothers are more likely to have frequent contact with
grandchildren than are paternal grandmothers (52%
vs 38%). On the other hand, there is no significant
difference between maternal and paternal grandfathers in frequent visiting, and the difference between paternal grandmothers and grandfathers is
small. It appears tnat having a daughter as the link to
the grandchildren is a salient factor for grandmothers
but not grandfathers. The same general pattern exists
when the focus is on infrequent visiting—maternal
grandmothers are least likely to visit infrequently
with a set of grandchildren, paternal grandfathers are
most likely to visit infrequently. A reasonable concluVol. 38, No. 3,1998
281
sion is that the role of gender and lineage in predicting grandparent—grandchild interaction should be
neither neglected nor exaggerated.
Multivariate Analysis
To explore the effects of the independent variables
on frequency of grandparent-grandchild set contact
in a multivariate analysis, we ran separate logistic
regressions to predict frequent and infrequent contact for grandfathers and grandmothers. The results of
these four analyses are shown in Table 2, where odds
ratios are presented. All of the odds ratios shown in
the table are statistically significant at .05. These results confirm most of our expectations regarding the
effect of various factors or frequency of grandparent
contact with a particular set of grandchildren.
The two key variables related to opportunity for
frequent contact come through strongly. Geographic
proximity is the most powerful predictor of either frequent or infrequent contact, as every previous study
also has reported. Number of grandchild sets is also
significant in each analysis shown in Table 2. Living
further away and having multiple sets of grandchildren both predict that grandparents are less
likely to have frequent contact and are more likely
to have infrequent contact with their grandchildren.
Consistent with findings reported by Cherlin and
Furstenberg (1986), the other two indicators of opportunity do not have any significant effects. Having
Table 2. Odds Ratios3 from Logistic Regressions Predicting
Frequent and Infrequent Visiting1' Between Grandparent
and Grandchild Set, by Sex of Grandparent
White grandmothers. Educational level was not a
significant predictor of frequent contact but was significant for infrequent contact. Other things being
equal, grandparents with higher education are less
likely to have infrequent contact with their grandchildren. The most straightforward interpretation of this
finding is that those with higher socioeconomic status
are more likely to have the necessary resources to overcome obstacles to visiting with children and grandchildren.
A central interest of this study concerns the effects
of marital status, lineage, and affect on contact between grandparents and grandchild sets. In general,
each of these variables operates in the expected direction. Consistent with previous studies, a more positive relationship between the grandparent and his/her
child is related to greater likelihood of frequent contact with the grandchild set (and less likely infrequent
contact). Married grandparents tend to have more
Dependent Variable
Grandfathers
Independent
Variable
G1 Age
Maternal lineage
Black
Other
In (Distance)
G1 Education
C1 Works
Grandchild sets
Frequent
Visiting
Infrequent
Visiting
0.664
0.334
Grandmothers
Frequent Infrequent
Visiting
Visiting
0.977
1.728
2.677
0.879
0.334
1.226
0.852
0.627
2.561
0.868
0.786
0.381
2.046
0.257
0.311
6.610
2.795
0.568
0.671
1.987
freauent contact than their nonmarried counterparts.
2.076
1.150
62.061
0.750
0.026
1.338
27.221
0.658
0.102
Pseudo R2
-2LogL
% Predicted
correctly
% Prevalence
0.449
0.483
929
689
0.460
1952
0.438
1090
82.1
62.2
89.9
82.9
83.7
52.6
90.0
86.0
N
1560
1560
3069
3069
Ana maternal grandparents are more likely to have
frequent contact than paternal ones. In addition, the
expectation that these variables differ by grandparents'
gender is confirmed.
Marital status is consistently of greater importance
in predicting contact with grandchildren for grandfathers than it is for grandmothers. In fact, widowed
grandmothers are not significantly less likely than married ones to have frequent contact. Grandfathers who
are divorced, separated, or remarried are especially
unlikely, compared to their married counterparts, to
maintain frequent visiting patterns with their grandchildren. The importance of lineage, on the other hand,
is greater for grandmothers than grandfathers. Although maternal grandmothers are much more likely
than paternal grandmothers to have frequent contact
with their grandchildren, maternal and paternal
grandfathers are not significantly different from each
other. Also, the quality of the relationship with a child
is a stronger predictor of frequent contact with a
grandchild for grandmothers than grandfathers. The
magnitude of the combined effect of gender, marital
status, and lineage on frequency of contact can be
made clearer by comparing predicted probabilities
that grandparents who differ on selected combinations
of these variables have frequent (or infrequent) contact with any particular grandchild set. Using the
logistic regression equations, we predict the probability of frequent (or infrequent) contact when all variables except marital status and lineage are set at
baseline values (baseline values, except for distance,
reflect sample means or modes). Specifically, we predict the probability of frequent contact for grandmothers and grandfathers with different combinations of lineage and marital status when they have
the following characteristics: age = 62 years; race =
White; education = 12 years; health = good; relationship with child = excellent; distance = 30 miles;
number of grandchild sets = 2; and grandparent is
not in the labor force. The baseline values for predicting infrequent contact are identical, except that distance is set at 500 miles. Results of this exercise are
shown in Table 3.
C1 Widow
1.149
2.061
G1 Separated/
divorced
G1 Remarried
G1 Poor health
G1-C2 affect
Constant
Note: Significance tests, based on Huber standard errors, were
calculated to correct for presence of multiple grandchild sets with
same grandparent in sample.
"Odds ratios that are not statistically significant at the .05 level
are omitted from this table. All of the odds ratios reported in this
table are significant at the .05 level.
b
Frequent means once per week or more; infrequent means
once per year or less.
poor health or being in the labor force do not appear
to interfere with the ability of grandparents to visit
with grandchildren. (Note, however, that grandparents
living in institutions are not included in this sample.)
The sociodemographic variables of age, race, and
education are included in the analyses, although no
particular expectations regarding tneir effects were
anticipated. An effect of age (net of health, marital
status, and other factors) is observed only for frequent
contact by grandmothers. In this one case there is evidence of advanced age being associated with a lower
likelihood of frequent contact. Perhaps in this case age
is associated with greater difficulty in transportation,
but the total effect is not very great. The absence of
racial difference in frequency of contact suggests that
cultural difference in grandparenthood between Whites
and Blacks may not be significant. This conclusion,
however, needs to be qualified by the fact that Black
grandmothers are much more likely than White ones
to be coresiding with one set of grandchildren. (Cases
of coresidence between generations were excluded
from the analysis.) In the NSFH2 sample, 30% of Black
grandmothers lived in a household that contained one
or more sets of grandchildren, compared to 9% of
282
The Gerontologist
Table 3. Predicted Probabilities of Grandparents Having Frequent and Infrequent Contact With Any Particular Grandchild Set
by Grandparents' Sex, Marital Status, and Lineage, Under Baseline Conditions"
Predicted Probability of Frequent Contact
Grandmothers
Measure
Married
Widowed
Sep/Div/Remar
Predicted Probability of Infrequent <Contact
Grandmothers
Grandfathers
Grandfathers
Maternal
Paternal
Maternal
Paternal
Maternal
Paternal
Maternal
Paternal
0.530
0.452
0.408
0.395
0.324
0.286
0.418
0.215
0.172
0.385
0.193
0.153
0.148
0.264
0.260
0.217
0.364
0.359
0.161
0.278
0.419
0.224
0.366
0.520
Note: Sep/Div/Remar = separated/divorced/remarried.
"Baseline conditions are described in the text.
The effect of marital status on grandfathers' visiting
patterns with grandchildren is striking. Other things
being equal, married grandfathers are twice as likely
as widowed ones to see their grandchildren frequently
and about half as likely to see them on an infrequent
basis. Divorced, separated and remarried grandfathers
are even less likely to have frequent contact with grandchildren. When grandfathers do not have a spouse to
assist them in maintaining kinship ties, they are much
less likely to experience high levels of interaction with
their grandchildren. A considerably smaller effect of
marital status is observed for grandmothers: divorced
and separated grandmothers are about 75% as likely
to visit frequently with grandchildren as married ones.
The stronger bond typical between a mother and
daughter than between a mother and daughter-inlaw is reflected in lineage differences in predicted frequency of contact between grandmothers and their
grandchild sets. Under the conditions specified above,
53% of maternal married grandmothers are predicted
to have frequent contact, compared to 40% of paternal
ones. Among grandfathers, it makes much less difference whether the link to their grandchildren is through
a son or a daughter. In combination, gender, marital
status, and lineage are found to sharply differentiate the
experience of grandparents. Comparing the extreme
categories in Table 3, one sees that maternal married
grandmothers are 3.5 times as likely to have frequent
contact with their grandchildren as paternal, separated,
or divorced grandfathers (.530 vs .153). The contrast
between these two categories in likelihood of having
infrequent contact is equally dramatic (.148 vs .520).
Conclusion
In conclusion, we first note that, based on a large
national sample of grandparent-grandchild set dyads,
six factors stand out as important predictors or frequent or infrequent contact:
1. Geographic distance is the strongest predictor
of frequent or infrequent contact. All relevant studies
have found this result, and it is evident that most grandparents maintain frequent contact with their grandchildren when it is feasible to do so.
2. The quality of the relationship between a grandparent and his/ner child affects the quantity of con-
Vol. 38, No. 3,1998
283
tact between the grandparent and that set of grandchildren. As reported by other studies, the middle generation may act as a gatekeeper for the grandparentgrandchild relationship.
3. The number of grandchild sets makes a difference. As the number of sets increases, the likelihood
of frequent contact with any particular set decreases
and the likelihood of infrequent contact increases.
Nevertheless, the probability of having frequent contact with at least one set of grandchildren is greater
for grandparents with multiple sets than for those with
only one set.
4. Gender of the grandparent is relevant. Grandmothers are more likely than grandfathers to have frequent contact and are less likely to have infrequent
contact.
5. It makes a difference if grandparents are linked
to a set of grandchildren through a daughter or a son.
Maternal grandparents are more likely than paternal
grandparents to have frequent contact with sets of
grandchildren and less likely to have infrequent contact. The effect of lineage is greater for grandmothers
than grandfathers.
6. Grandparents' marital status has a significant
effect on the likelihood of frequent or infrequent contact with grandchildren. This effect is especially strong
for grandfathers. For both grandmothers and grandfathers, the ordering from most to least likely to have
frequent contact is married, widowed, remarried, divorced.
The most important conclusion to draw from this
analysis is that grandparenting needs to be examined
within a complete kinship framework where gender
matters. Starting with G1, it makes a difference whether
the grandparent is a grandmother or a grandfather. It
makes a difference whether or not the grandparent
has experienced marital disruption. It makes a difference how many grandchild sets the grandparent
has. And it makes a difference whether the grandparent is linked to grandchildren through a son or a
daughter. In other words, general discussions of grandparenting are of limited usefulness. The relationship
between a grandparent and a grandchild is always
located within a particular kinship network. Differences
in kinship networks help to explain differences in
grandparenting behaviors.
This study overcomes a limitation of most previous
grandparenting studies by having a large, nationally
representative sample of grandparents. There are,
nevertheless, three significant limitations related to
using these data from NSFH2. The first, apparent to
a perceptive reader, is that contact between grandparents and grandchild sets is not measured directly.
As discussed in the Appendix, there is good evidence
that frequency of contact between G1 and G2 is a
reasonable proxy for frequency of contact for the corresponding G1-G3 relationship, but it would be much
better to measure this directly. A second limitation is
that information on characteristics of the individual
grandchildren is not available. The third limitation is
that consequences of contact frequency cannot adequately be assessed. Hopefully more complete information on the grandparent-grandchild relationship
will be collected in future surveys so that the findings
suggested by this analysis can be examined more rigorously and their significance explored more fully.
As the literature on intergenerational relationships,
including grandparent-grandchild relationships, grows,
we are gaining new insights into the role of kin ties in
the aging experience. It is not surprising that these
new insights stimulate new questions that require
additional research. To deepen our understanding of
grandparenting, future studies might explore the following questions: What are the implications of changing family gender roles for relationships between
generations? Although grandfathers in aggregate have
different relationships with grandchildren than grandmothers, it is obvious that there is great diversity among
both grandfathers and grandmothers. Some grandfathers are highly involved with their grandchildren;
some grandmothers are marginally involved with their
grandchildren. Do fathers who play a more active role
in caring for their young children later develop into
grandfathers who are more engaged with their grandchildren? Do mothers who turn over much of the parenting role to others develop into less engaged grandmothers?
Greater attention needs to be given to how grandparents relate to different sets of grandchildren and
to different grandchildren within a set. The first step,
of course, is to directly measure grandparent contact
and relationship with each set of grandchildren and
with each grandchild in each set. Then the question
of what leads to differential involvement can be explored. In addition to factors suggested by findings of
this study, attention should be given to the relationship between the grandparent and the spouse of
the child who links him or her to the grandchildren.
It will be important to examine how the needs of various grandchild sets affect the relative share of the
grandparents' attention that they receive. And characteristics that differentiate grandparental attention
among siblings in a set could be explored.
As the study of intergenerational relationships progresses, it becomes increasingly clear that relationships develop over time. Cross-sectional studies that
attempt to explain grandparent-grandchild relationships
by focusing on characteristics at one point in time are
simply inadequate. New efforts should be made to
284
include information on the process through which
relationships develop and change. Well designed longitudinal studies are ideal for this purpose, but it also
is possible to collect retrospective life course histories
that shed light on the development of relationships.
The G1-G2 relationship, which plays such an important role in the G1-G3 relationship, has a history.
Are there critical events in the parent-child relationship that later pull grandparents and grandchildren
together, or which drive them apart? Exploration of
these issues must await the collection of appropriate
data.
Finally, the consequences of differing levels of mutual engagement of grandparent and grandchildren
need to be examined more carefully. Are there direct
benefits to both generations that come from maintaining frequent contact and high emotional attachment?
Beyond the individual level, it also is worth asking how
grandparent-grandchild connections might benefit
the larger society. A number of social changes over
the past century have tended to increase age segregation in our society: changes in schools, work, churches,
age-graded organizations, retirement, and extended
family living arrangements. The most significant linkages across age groups probably occur because of kinship ties. How important are grandparent-grandchild
connections in preserving age integration? And what
are the factors that enhance the quantity and quality
of these intergenerational relationships?
References
Aldous, I. (1995). New views of grandparents in intergenerational context. Journal of Family Issues, 76, 104-122.
Cherlin, A. I., & Furstenberg, F. F. (1986). The new American grandparent: A place in the family, a life apart. New York: Basic Books.
Cooney, T. M., & Uhlenberg, P. (1990). The role of divorce in men's
relations with their adult children after midlife. Journal of Marriage
and the Family, 52, 677-688.
Fischer, C. S. (1982). The dispersion of kinship ties in modern society.
Journal of Family History, 7, 353-375.
Cravenish, B. A., & Thomson, E. (1996). Marital disruptions and grandparent relationships (NSFH Working Paper No. 74). Madison: University of Wisconsin, Center for Demography and Ecology.
Hagestad, C. O. (1986). The aging society as a context for family life.
Daedalus, 775, 119-139.
Harris, L, & Associates. (1975). The myth and reality of aging in America.
Washington, DC: National Council on Aging.
Hernandez, D. J. (1993). America's children: Resourcesfromfamily, government
and the economy. New York: Russell Sage Foundation.
Hodgson, L. C. (1992). Adult grandchildren and their grandparents: The
enduring bond. International Journal of Aging and Human Development, 34, 209-224.
Johnson, C. L. (1985). Crandparenting options in divorcing families: An
anthropological perspective. In V. L. Bengtson & I. F. Robertson (Eds.),
Crandparenthood (pp. 81-96). Beverly Hills, CA: Sage.
King, V., & Eider, C. H. (1995). American children view their grandparents: Linked lives across three rural generations. Journal of Marriage
and the Family, 57, 165-178.
Matthews, S. H., & Sprey, J. (1985). Adolescents' relationships with grandparents: An empirical contribution to conceptual clarification. Journal
of Gerontology, 40, 621-626.
Roberto, K A., & Stroes, J. (1992). Grandchildren and grandparents: Roles,
influences, and relationships. International Journal of Aging and Human Development, 34, 227-239.
Robertson, I. F. (1995). Crandparenting in an era of rapid change. In
R. Blieszner & V. H. Bedford (Eds.), Handbook on aging and the
family (pp. 243-260). Westport, CT: Greenwood Press.
Rossi, A. S. (1995). Wanted: Alternative theory and analysis modes. In
V. L Bengtson, K. W. Shaia, & L. M. Burton (Eds.), Adult intergenerational
relations: Effects of societal change (pp. 264-276). New York: Springer.
Rossi, A. S., & Rossi, P. H. (1990). Of human bonding: Parent-child
relations across the life course. New York: Aldine de Grayter.
Szinovacz, M. (1998). Grandparents today: A demographic profile. The
Cerontologist, 38, 37-52.
The Gerontologist
Uhlenberg, P., & Kirby, J. B. (in press). Crandparenthood over time:
Historical and demographic trends. In M. Szinovacz (Ed.), Handbook
on grandparenthood. Westport, CT: Greenwood Press.
Whitbeck, L. B., Hoyt, D. R., & Huck, S. M. (1993). Family relationship
history, contemporary parent-grandparent relationship quality, and the
grandparent-grandchild relationship. Journal of Marriage and the Family,
55, 1025-1035.
Received February 18, 1997
Accepted July 29, 1997
Appendix
To assess how well frequency of G1-G2 contact serves
as a proxy for frequency of C1-G3 contact, we examine
cases where C1 reports only one child with children in
his/her household. The respondents (G1) report how frequently they have contact with each child and with their
grandchildren (both on a scale from 1 to 6). When there
is only one set of grandchildren, we expect that in most
cases the contact reported with grandchildren will be the
frequency of contact they have with the grandchild set
of the child who has children in his/her household. (The
exceptions would be when grandparents have other grandchildren not living in their parents' household. Thus some
grandparents might be reporting frequency of contact
with a grandchild who is not in the set of interest. We
expect the results would be even stronger if there were a
way to exclude these cases.) As reported in the text, in
86% of the cases the frequency of contact with the child
and with grandchildren is either identical or in an adjacent category.
To examine how robust the results of Table 2 are, we
ran the logistic regressions restricted to cases where there
is just one set of grandchildren (n = 1,243). We compare the results of this regression model when the dependent variable is frequent (or infrequent) contact be-
Vol. 38, No. 3, 1998
285
tween G1-G2 (as in Table 2) with the results when the
dependent variable is frequent (or infrequent) contact
between G1-G3. In this analysis we must, of course, drop
number of grandchild sets as an independent variable.
In addition, we have made marital status of the grandparent a dichotomous variable (married to other parent
of G2 vs other), and we have included gender of the
grandparent as an independent variable. The results are
shown in Table 4 for the variables of interest.
Overall, the results presented in Table 4 suggest that
using G1-G2 contact as a proxy for G1-G3 is not problematic. The only exception might be with the effect of
maternal lineage. It is plausible, however, that lineage is
not significant in the special case where grandparents have
only one set of grandchildren. Paternal grandparents with
only one set of grandchildren will not see any grandchildren if they do not see their son's grandchildren.
Table 4. Regressions Predicting Frequent and Infrequent
Contact When Grandparents Has Only One Grandchild Set
Frequent
Contact
Infrequent
Contact
Dependent Variable
G1-G2
G1-G3
G1-G2
C1-G3
Maternal lineage
In (distance)
G1 not married3
G1-G2 affect
G1 male
1.616*
0.281*
0.471
1.234
0.517*
0.689
1.232*
0.601
0.732
2.440*
2.903*
0.694*
1.049
1.108
2.010*
2.013*
0.699*
1.726
1.364*
0.491*
*Significant at .05.
G1 not married to other parent of G2.
a