Content uploaded by Warren Miller
Author content
All content in this area was uploaded by Warren Miller on Aug 03, 2015
Content may be subject to copyright.
Disagreement between Spouses’ Childbearing Desires:
Changes during Five Years of Marriage
Warren B. Miller
Transnational Family Research Institute
Aptos, CA
David J. Pasta
Ovations Corporation
San Francisco, CA
2006
Acknowledgements
Supported by Grant R01 HD23900 from the National Institutes of Child Health and
Human Development.
1
Disagreement between Spouses’ Childbearing Desires:
Changes during Five Years of Marriage
Abstract
Disagreement about childbearing has been shown to affect fertility, yet there is
little systematic knowledge about how such disagreement changes in the course of
marriage. We describe a theoretical framework that includes the childbearing desires and
perceived spouse childbearing desires of both partners. Based on data collected from 389
married couples at five points in time during a five year period, we use hierarchical linear
modeling to examine how initial (time-invariant) characteristics of each spouse and the
couple itself and certain follow-up (time-varying) events experienced by them affected
disagreement in their childbearing desires. Separate analyses are conducted on couples
who have children during the follow-up and those who do not. Results indicate that
disagreement increases among those having children and remains relatively stable among
those who do not. The predictors of change in disagreement indicate some dynamics that
underlie both the antecedents and consequences of couple disagreement about
childbearing.
Key words: childbearing desires/couple disagreement/hierarchical linear model
Disagreement between Spouses’ Childbearing Desires:
2
Changes during Five Years of Marriage
Introduction
Available evidence from national studies in the United States indicates that many
spouses disagree on the number of children that they desire (Thomson, McDonald, &
Bumpass, 1990) and it seems reasonable to assume that there are similar degrees of
disagreement in other motivational antecedents to childbearing. Such disagreements
have practical consequences because differences in spousal fertility desires (Thomson,
McDonald, & Bumpass, 1990; Thomson, 1997) or intentions (Miller & Pasta, 1995a;
Thomson, 1997) have been shown to effect the rate of childbearing. For this and other
reasons, researchers have begun to develop various models of couple motivation and to
explore how well they explain fertility behavior (Thomson, 1983; Bagozzi & Van Loo,
1987; Miller & Pasta, 1995a; Miller, Severy, & Pasta, 2004).
In this paper we consider one aspect of couple fertility motivation, namely
differences between the two members of a couple. Because childbearing desires
represent the simplest and best measure of individual fertility motivation, we focus on
marital partner differences in those desires. Specifically, we want to look at variation in
the magnitude of childbearing desires across couples and how that magnitude varies over
time and in response to fertility events such as pregnancy and the birth of a child. Our
approach to these questions is guided by a recently developed theoretical framework
(Miller, Severy, & Pasta, 2004) that examines the critical issues that are involved when
the motivations of two individuals produce a conjoint behavioral outcome. Our findings
will help to clarify the important couple-level phenomenon of motivational disagreement,
which is known to have significant consequences for fertility and is certainly of
considerable importance for the understanding of all intimate personal relationships.
3
Theoretical Framework
Dyads in various forms are of fundamental importance to all human survival and
reproduction. Whether it is the parent/child dyad, the dyadic relationships between
playmates and between workmates, or the mating dyad, this form of intimate relationship
exists because many of the adaptive problems that we faced during our evolution as a
species were more successfully solved through the joint behavior of two individuals
working together (Miller & Rodgers, 2001). However, dyads face a critical problem,
namely how to unify the separate interests of each individual in a way that satisfies both
of them. Put in terms of our specific focus here, the problem becomes how are the
separate fertility motivations of two spouses combined in a way that produces their
conjoint behavior and, ultimately, their actual joint fertility. We address this problem
with a theoretical framework (Miller, Severy, & Pasta, 2004) that integrates a model of
individual fertility motivation with a theory-of-mind-based model of dyadic interaction
and mutual perception.
The motivational antecedents to childbearing and the behavior that follows from it
may be thought of in terms of a traits-desires-intentions-behavior (T-D-I-B) sequence.
Motivational traits are enduring dispositions that direct individual behavior toward
certain more or less broad goals. Miller (1995) has described two broad motivational
traits relevant to fertility, namely positive and negative childbearing motivation. Desires
represent psychological states that express what an individual wants to do with respect to
specific goals. They derive from motivational traits but are also influenced by attitudes,
beliefs, and other trait-like factors. With respect to the domain of fertility behavior, there
are three important desires, namely childbearing desires, child-number desires, and child-
timing desires (Miller, 1994). Intentions represent psychological states that express what
an individual actually plans to do with respect to the same specific goals. They are
derived from desires and are shaped by reality constraints, in particular by partner
desires. In the fertility domain, there are three types of intentions that correspond to the
three types of desires. Instrumental behavior represents what the individual actually does
in order to accomplish his or her specific goals. The type of behavior enacted by the
individual has a major role in whether or not a particular fertility event such as pregnancy
4
is achieved or avoided. In previous work, we showed that child-timing intentions and
childbearing intentions were the strongest predictors of proceptive behaviors (efforts to
achieve conception), depending in part upon whether the behavioral outcome variable
was time-ordered or not (Miller & Pasta, 1995a). In subsequent work, we showed that
contraceptive behaviors (efforts to prevent conception) were also well predicted by
fertility intentions (Miller & Pasta, 1996a).
The T-D-I-B sequence applies to individuals but because the fertility domain
typically involves the intimate interaction of two people, it is necessary to elaborate the
sequence to include both partners. We do this by incorporating into our model the
theory-of-mind capability that humans possess (Premack & Woodruff, 1978), especially
our ability to apprehend others’ emotional and motivational states (Eisenberg & Miller,
1987). Figure 1 shows the desires-intentions-behavior part of the sequence as applied to
a married couple. The translation of desires into intentions for each spouse is shown to
be influenced by each individual's perception of their spouse's desires. Their separate
intentions are then implemented in the course of enacting joint behavior in the form of
proception or contraception. Note that the translation step involves integration within
each individual, although it is obviously based on multiple interactions between spouses,
whereas the implementation step involves integration between the two individuals. In a
previous paper (Miller & Pasta, 1996b), we have
5
examined how the integrations involved in both of these steps varies by sex and parity
and across time.
In this paper we are broadly interested in the desires and perceived desires of
married couples, how they interact over time, and what factors affect that interaction.
Because childbearing desires (i.e., the desire for a[nother] child) are in many respects the
most fundamental of the three fertility desires (Miller, 1994; Miller & Pasta, 1995a), we
will focus on those. In terms of Figure 1, we will limit our consideration to the
desires/perceived spouse desires part of the sequence. By using only that part and
rotating the figure clockwise 90°, we can visualize each spouse's childbearing desires and
their perception of the other's desires in a horizontal row as shown in Figure 2.
Husband’s
Desires
Perceived
Wife’s Desires
Perceived
Husband’s Desires
Wife’s
Desires
Husband’s
Intentions
Wife’s
Intentions
Conjoint
Behavior
Figure 1. A Schematic Representation of the Desires-Intentions-Behavior part of
the TDIB Sequence As Applied to Married Couples.
6
The notation used in this figure is as follows: Wa and Ha indicate wife and
husband actual childbearing desires and are surrounded by solid-line circles; Hp indicates
the husband's childbearing desires as perceived by the wife and is surrounded by a
dashed-line circle. Wp, the wife's childbearing desires as perceived by the husband, is
omitted for the sake of simplicity but its location is shown by a faint dashed-line circle.
We have drawn a rectangle around the circles representing Wa and Hp and around those
representing Ha and Wp to indicate that former two are a part of the wife's consciousness
and the latter two are a part of the husband's consciousness.
Figure 2 also shows two sets of arrows: three curved, double-headed arrows,
numbered 1, 2, and 3, and four straight, single headed arrows, numbered 4, 5, 6, and 7.
The first set of arrows indicate correspondences or correlations between the two elements
indicated by each arrow, at least one of which in each case is based in the wife’s
consciousness. Arrow 1 indicates the correspondence between wife's actual childbearing
desires and husband's actual childbearing desires and thus represents actual agreement
between partners. Arrow 2 indicates the correspondence between wife's actual
childbearing desires and her perception of her husband's childbearing desires and thus
represents the wife's perceived agreement. Arrow 3 indicates the correspondence
between the wife’s perception of the husband's childbearing desires and the husband's
actual childbearing desires and thus represents the wife's accuracy of perception. In order
to avoid crowding the figure, we have omitted the arrows representing the husband's
perceived agreement and the husband's accuracy of perception. The second set
of arrows shown in Figure 2 represent influence processes that involve Hp and affect the
wife’s three correspondence states. The first process, represented by arrow 4, is
acceptance. In it, the wife's perception of her
7
WW
WW
W
Wa Hp Ha
1
4
5
6
7
2
3
Wife’s Consciousness Husband’s Consciousness
Figure 2. A Schematic Representation of Wife’s and Husband’s Actual Desires
(Wa and Ha) and Wife’s Perception of Husband’s Desires (Hp), with Three
Correspondence States Indicated by Double-Headed Arrows 1, 2, and 3 and Four
Influence Processes Indicated by Arrows 4, 5, 6, and 7.
8
husband's desires (Hp) influences her own desires (Wa). The second process,
represented by arrow 5, is attribution. In it, the wife's desires (Wa) influence her
perception of her husband's desires (Hp). The third process, represented by arrow 6, is
accommodation. In it, the wife's perception of the husband's desires (Hp) influences his
actual desires (Ha). Finally, the fourth process, represented by arrow 7, is apprehension
(as in "to apprehend," meaning to understand intuitively, to correctly perceive). In it,
husband's actual desires (Ha) influence the wife's perception of his desires (Hp). Again
in order to avoid crowding the figure, we have omitted the four influence processes that
involve Wp and affect the husband’s three correspondence states. Obviously, there are a
number of other influence processes at the psychological, dyadic, and social levels that
may also affect childbearing desires and perceived desires. In order to avoid excessive
complexity, these are also not included in the figure.
We stated above that the four intra-couple influence processes just described
affect the wife’s three correspondence states of actual agreement, perceived agreement,
and accuracy of perception. Figure 3 illustrates the action of each of these processes by
showing how a change in one component of our Wa-Hp-Ha schematic along a
hypothetical child-bearing dimension measured from 0 to 20 affects certain
correspondence states, assuming that the other two components are held constant. We
realize that in reality all components of our model, not just one, are regularly undergoing
both systematic and random change and, therefore, that Figure 3 represents four
simplified scenarios. However, simplification serves the heuristic purpose of
demonstrating the differences among the four interactive processes and the three
correspondence states, while simultaneously indicating how each process, when acting
alone, affects certain states.
The first row of Figure 3 shows that when apprehension increases and moves Hp
toward Ha, then accuracy of perception increases and perceived agreement decreases; and
the second row shows that when attribution increases and moves Hp toward Wa, then
9
accuracy of perception decreases and perceived agreement increases. In other words,
everything else (specifically Wa and Ha) being constant, these two processes have exactly
opposite effects. The story is not so simple when it comes to the other two influence
processes. The third row shows that when accommodation increases and moves Ha
toward Hp, then actual agreement increases and so does accuracy of perception; and the
fourth row shows that when acceptance increases and moves Wa toward Hp, then again
Influence ___Actual and Perceived Desires___ Correspondence
__ Processes___ ____States____
Wa Hp Ha
Apprehension Accuracy of Perception
Perceived Agreement
Attribution Accuracy of Perception
Perceived Agreement
Accommodation Actual Agreement
Accuracy of Perception
Acceptance Actual Agreement
Perceived Agreement
_____________________________________________
0 10 20
Figure 3. A Schematic Representation Showing How Each of Four Processes that
Influence the Wife’s Actual Childbearing Desires (Wa), the Perceived Childbearing
of her Husband (Hp), and the Husband’s Actual Childbearing Desires (Ha) Affect the
Correspondence between them, Holding Other Factors Constant.
10
actual agreement increases but now so does perceived agreement. In other words, these
two processes have in part the same effects and in part different (but not opposite) effects.
What happens in the real world when one of these influence processes changes?
We can think about this question with the help of a specific example where Wa, Hp, and
Ha have values on the childbearing desires scale shown in Figure 3. Suppose that W and
H have a talk about childbearing, the first one for a few years. Suppose also that during
the talk W discovers that her perception of H's childbearing desires has been quite wrong,
and that whereas she thought he was a 9 before the talk, afterwards she concluded he was
a 12 (still some bias here because he is actually a 16). This change in her perception
might well set in motion some changes in her own desires. For example, she might
accept his point of view enough so that over time she moved from being a 5 to being an
8. On the other hand, her recognition of the magnitude of their disagreement might result
in her feeling hopeless about agreement and lead her to reject his point of view by
moving to be a 3 or a 2.
Suppose additionally that H changes jobs to one where he will be traveling less
and will be able to devote more of his time to childcare. Suppose also that this generates
a further increased acceptance of H's childbearing desires by W. As she moves from
being a 8 toward being, say, a 12, we can imagine that the additional discussions likely to
be stimulated by these changes might well alter her perception of H's childbearing desires
toward greater accuracy, causing the balance between her attribution and apprehension to
equilibrate in some new position, such as 14, making her perception closer to H's reality.
In other words, W's increased acceptance of H's childbearing desires as a result of his
greater availability for childcare leads her to a more accurate perception of his desires,
although some attribution continues to affect her perception.
It is apparent that the model we have elaborated here –illustrated in Figure
2 by the Wa-Hp-Ha schematic with its four connecting process arrows-
11
is highly recursive (i.e., with variable effects looping back on themselves). This is even
more the case if the husband part of the model (the Ha-Wp-Wa schematic) is added to
make the whole framework more completely dyadic. Given the amount of model
endogeneity, separation of the four processes of acceptance, accommodation, attribution,
and apprehension during data analysis is certain to be extremely difficult, if not
impossible. Nonetheless, with a few simplifying assumptions, some progress can be
made. It will be noted in Figure 3, for example, that the two processes of apprehension
and attribution are related in that both affect the position of Hp relative to Wa and Ha and
the two have opposite effects on accuracy of perception and perceived agreement. On the
other hand, the two processes of accommodation and acceptance are related in that both
move Wa and Ha relative to each other and the two have the same effect on actual
agreement. These pair-wise similarities have led us to two separate analyses. In this
paper, we examine couple actual agreement, or more accurately, degree of
agreement/disagreement. As a result, we interpret our findings in terms of
accommodation and acceptance. No effort will be made to tease apart the effects of these
two processes, although we believe that acceptance is probably a more important process
than accommodation, given the strength of individualism in the U.S. culture. In a
companion paper to this one (Miller & Pasta, 2004), we examine bias in perception and
interpret our findings in terms of apprehension and attribution.
There is one final aspect of our theoretical framework that requires discussion.
Figure 4 shows some of the ways that Ha can change with respect to Wa. Such
movement represents the net effect of acceptance and accommodation and results in a
change in the amount of actual agreement. Suppose that Ha begins at position 1 and that
it changes over time to positions 2, 3 or 4. The change to position 2 represents a change
in the direction of disagreement; the change to position 3 represents a change in the
magnitude of disagreement; and the change to position 4 represents a change in both the
direction and magnitude of disagreement. Although both direction and magnitude are of
12
interest in the study of disagreement, we focus here on magnitude because we have found
it to be the more important component when studying marital dyadic behavior (Miller &
Pasta, 1995a). In the data analyses reported below, we use the absolute difference
between Wa and Ha to examine the magnitude of couple disagreement.
__________________________________________________________________
0 10 20
Figure 4. A Schematic Representation of Different Possible Positions of Ha Relative to Wa
Along a Single Dimension of Childbearing Desires.
Methods
Sampling, Data Collection, and Variable Construction
The data analyzed in this paper were collected during a longitudinal research
project concerned with childbearing motivation and its effects on fertility in 201 married
couples with no children and 200 married couples with one child at the beginning of the
study. All of these couples were living in a largely urban and suburban county located in
the San Francisco Bay Area. Sampling was conducted by telephone using a modified
area probability technique. Eligible couples included those who were currently married
with zero or one child, in which both spouses spoke and read fluent English, neither
13
Ha Ha
Wa HaHa
4 2 1 3
spouse was sterilized, and the wife was not pregnant and was age 18 through 39. The
average age of the husbands and wives was 31.6 and 29.6, respectively. Marital duration
for the parity-zero couples was just over 3 years and for the parity-one couples was over
four and one-half years. The average husband had 16.7 years of education; the average
wife had 15.9 years. Couple income averaged $71,000 in 1989. Additional
characteristics were the following: 7% of the males, 13% of the parity-zero females and
35% of the parity-one females were not currently employed; 25% of the sample was
Roman Catholic, 44% was Protestant, and 21% had no religion; and 81% of the sample
was white, 7% was Asian, 5% Latin, and 2% black. Further information about sampling
and social-demographic characteristics of the sample is available in other publications
(Miller, 1992, 1994; Miller & Pasta, 1993).
Data were collected from this sample during five structured interviews: an initial
interview and four follow-up interviews 12, 24, 42, and 60 months later. At the initial
interview each husband and wife was interviewed separately and privately, either at the
investigator's research offices or, more commonly, in the couple's own home. At the
same time, each respondent was administered several self-report inventories. The four
follow-up interviews were conducted by telephone and the inventories were mailed to
them with a return envelope, helping to maintain a low loss to follow-up. Because of the
need for dyadic data in the current study, we treated couples as lost to follow-up when
they had separated permanently or divorced. Of the 401 original couples, 12 were lost to
follow-up after the initial interview and are not included in this analysis. For the
remaining 389 couples, we have two years of data from 18 couples, three years from 47
couples, four years from 31 couples, and five years from 293 couples.
In preparation for data analysis, we selected from five categories of variables that
our previous research has shown to be important determinants of childbearing desires and
their changes over time (Miller, 1994; Miller & Pasta, 1995b). These categories included
social and demographic background variables, personality traits, motivational antecedents
14
to childbearing, gender role variables, and couple interactions that centered around
childbearing. We discuss each of these categories in turn.
Data for the following background variables relevant to the current study were
collected during the initial interview: sex, age, years of education, religion, religiosity,
number of marriages, marital duration, beginning gravidity, and beginning parity. At
each follow-up interview we determined whether the wife had become pregnant during
the previous interval and if so, how often and what the outcome(s) was (were), which for
this study we classified as a failed pregnancy, a live birth, or still (currently) pregnant.
We also constructed a follow-up parity variable at each follow-up which took into
account not only changes through births but changes through adoption and alterations in
custody arrangements. During the four follow-up interviews, 73 failed pregnancies were
reported, 155 women were pregnant at the time of the interview; and 291 live births were
reported. Parity at the end of study was zero for 87 couples, one for 127 couples, two for
155 couples, three for 19 couples, and four for one couple.1
Also during the initial interview five personality traits were measured, using
Jackson's (1984) Personality Research Form. These traits were Achievement, Affiliation,
Autonomy, Order, and Nurturance.
During each interview, we measured various aspects of the motivational
antecedents to childbearing. Most important for our present purposes were Childbearing
Desires (Miller, 1994). Respondents were asked to indicate on a five-point scale how
certain they were about their desire to have a(nother) child and then to rate on a ten point
scale how much they desired to have a(nother) child. A composite Childbearing Desires
variable (range 2-20, with higher scores indicating greater childbearing desires) was
constructed by adding the two scores together after recoding answers to the first question
(1=1, 2=3.25, 3=5.5, 4=7.75, 5=10) in order to correct for the different range in response
1In order to avoid a possible distorting outlier effect, the one couple with four children was excluded from
the analysis at the final interview when they had had their fourth child.
15
values. The mean score across all five years was 13.88. We then asked two similar
questions about respondents' perceptions of the certainty and magnitude of their spouses'
desires for a(nother) child and constructed in a similar way a Perceived Spouse
Childbearing Desires variable. The mean score across all five years was 13.81. The
reliability of these measures is high and the evidence for their validity is strong, as Miller
has discussed elsewhere (1994).
Additionally, at each interview we measured the respondents’ Child-Number
Desires (Miller & Pasta, 1993), Child-Timing Desires (Miller & Pasta, 1994), and
Childbearing Motivation (Miller, 1995). The latter includes two orthogonal primary
scales, Positive Childbearing Motivation (PCM) and Negative Childbearing Motivation
(NCM).
Finally, we collected data relevant to two separate but related categories of
variables, gender role and couple interaction. With one exception, these variables were
based on single interview questions and with two exceptions, the data were collected at
every interview. The gender role variables included two questions (at the initial
interview only) about which spouse had primary responsibility for making the decision
about whether to have a(nother) child and when to have it. Response categories were: 1.
wife only, 2. wife mostly, 3. wife and husband equally, 4. husband mostly, and
5. husband only. The mean for both questions was about 2.8, with only a few 1s and 5s.
Also considered as gender role variables were a question about current employment
status and one about work satisfaction. The latter was rated on a five point scale from 1.
very low to 5. very high. Those not currently employed were given their most recent
rating. The mean score across five years was about 4.
The couple interaction variables included two questions about how good the
communication had been "between the two of you" about whether to have a(nother) child
and about when to have a(nother) child. Response categories were: 1. very good, 2.
mostly good, 3. both good and bad, and 4. bad. Initial interview sample means for the
16
two questions were 1.6 and 1.5 respectively, with 3s and 4s together constituting only
about 10% of the responses. Two additional couple interaction variables were based on
two questions about the amount of disagreement between the respondent and his/her
spouse about whether to have a(nother) child and when to have it. Response categories
were: 1. none, 2. a little, 3. a moderate amount, and 4. a great deal. Initial interview
sample means for these two questions were 1.5 and 1.6 respectively. The last couple
interaction variable was called Marital Problems and was based on three questions about
marital satisfaction (from 1. very high to 5. very low), the presence of serious conflicts
(from 1. no to 3. yes, a lot), and whether the respondent had given consideration to
separation (from 1. no to 3. yes, several times). We used the first principal component of
these three questions for variable construction.
Data Analysis
We used hierarchical linear models of change (Bryk & Raudenbush, 1992) and
the HLM software (Bryk, Raudenbush, Seltzer, & Congden, 1988). In our analysis, we
wanted to look at the magnitude of the difference between the husband's and the wife's
childbearing desires as a measure of actual disagreement and ask questions about the
patterns and determinants of that disagreement. To what extent do different couples tend
to have different amounts of disagreement? Does disagreement generally rise or fall over
time? Does birth of a child tend to increase or decrease disagreement? How does the
average level of childbearing motivation affect disagreement? Does it matter for
disagreement which spouse is more motivated? And finally, what is the predictive power
of the demographic characteristics, personality, or gender role of either spouse, or of the
couple's interaction on the magnitude of disagreement?
These questions about the magnitude of disagreement are best answered by
directly modeling the disagreement. Thus our outcome variable in these analysis was the
absolute value of the difference between the actual childbearing desires reported by each
17
member of the couple. For convenience, we refer to this as "absolute SARA," for
"absolute value of Spouse Actual minus Respondent Actual." Because we are taking the
absolute value, this is the same as "absolute RASA," the absolute value of Respondent
Actual minus Spouse Actual, and each couple contributes only one set of (up to) five
observations over time. Our initial Level 1 model included both time and parity:
(Absolute SARA)it = π0i + π1i (Time)it + π2i (Parity)it + eit
where i indexes the couples and t indexes the data collection time points.
Construction of Analysis Variables
In preparation for the analyses, we created mean variables (average of husband
and wife) and signed difference variables (husband minus wife) from many of the simple
(individual-based) variables already discussed. Exceptions were: 1. for background
variables common to the two spouses (e.g., marital duration, beginning parity) and for
fertility events occurring over the course of the study, a common couple-level variable
was constructed; and 2. for personality variables, the two individuals' variables were not
combined in any way and were simply designated as "husband" or "wife" versions. We
did not create any absolute difference predictor variables for these analyses. Although
such variables are of interest, we believed that including them in the current analyses
would produce a result that only showed how disagreement predicted disagreement and
would not allow us to draw conclusions about accommodation and acceptance.
Model Construction Criteria
In the analysis, our particular interest was in changes in absolute SARA over
time, so the initial model included time and parity as the only time-varying variables.
Having determined that there was sufficient variance in the parameter describing each
couple's across-time slope, we then added time-varying variables one at a time until a
time-related model was established. Then time-invariant predictors were added in a
18
forward-stepping manner to the second-level model. Guided largely by simple
correlations between residuals from the fitted second-level models and possible
explanatory variables, we systematically added variables to the second-level models.
Candidate variables were included in the base (intercept) model whenever they were
included in the model for a first-level parameter (slope). Some predictor variables were
added only to the base model. Because of the exploratory nature of our analyses, we
elected to be broadly inclusive in our p-value for retaining predictors, setting it at a two-
tailed p=0.10.
Results
We began our analysis by examining the time trend in absolute SARA (couple
disagreement). As expected, we found that time was a highly significant predictor of
absolute SARA and that there was ample parameter variance across couples. Based on
earlier research (Miller & Pasta, 1995b), we expected parity to be an important predictor.
We included average follow-up parity as a time-invariant value and found it was highly
significant. We then considered whether parity as a time-varying variable would add
predictive power, and we found that it did. Furthermore, we found that when parity was
included as a time-varying variable, the effect of time itself was greatly attenuated.
This led to reconsideration of the basic model for examining absolute SARA over
time. We recognized that only the couples who had changed parity over the five years
could contribute to the evaluation of parity as a time-varying variable, while all couples
could contribute to the evaluation of parity as a time-invariant variable. We performed
an analysis on only the couples who had changed parity over the five years (N=240
couples) and found that parity as a time-varying variable added significantly to the
prediction of absolute SARA even in the presence of time. Furthermore, with parity in
the analysis as a time-varying variable we found that time was no longer a statistically
19
significant predictor of absolute SARA. That is, for the parity changers we found that
disagreement changed across parity but did not change additionally across time.
We then considered the parity non-changers (N=149 couples). Initial parity
(which is equivalent to average follow-up parity for the non-changers) was included as a
time-invariant variable and time as the only time-varying variable. For this group, we
found that disagreement did change across time. Based on this fundamental difference in
the first-level model between the parity changers and the parity non-changers, we decided
to analyze the two groups separately.
Tables 1 and 2 present descriptive statistics by sex for the parity changer and non-
changer groups of all variables measured at baseline and used as predictors in the HLM
analyses. When we compared these two groups by sex for all these variables, we found
substantial differences, especially with respect to who had a mean initial score of 14.68 (p
<.001). For females the corresponding scores were 18.11 and 14.46 (p <.001). There
were similar very large differences
Table 1. Means and Standard Deviations for Husbands and Wives and for Husband-Wife
Differences and Husband/Wife Correlations for All Predictor Variables at Baseline
Among Parity Changers (N = 240).
Variable Husband Wife Husb-Wife
Husb/Wife
Mean S.D. Mean S.D. Mean S.D.
Correl.
Age 31.13 4.61 29.11 3.91 2.02 3.38
.696
Education 16.59 2.42 15.90 2.23 .69 2.05
.614
Protestant .46 .50 .46 .50 .00 .63.1
96
20
Roman Catholic .33 .47 .36 .48 -.03 .60
.221
Other Religion .07 .25 .07 .26 .00 .21
.642
No Religion .14 .35 .10 .31 .03 .44
.102
Religiosity 4.28 1.74 4.66 1.64 -.37 1.35
.683
Number of Marriages 1.10 .31 1.08 .29 .02 .39
.145
Marital Duration 3.71 2.64 3.70 2.64 .01 .26
.995
Gravidity .73 .88 .80 .97 -.06 .60.7
93
Parity .55 .50
Autonomy -1.28 2.83 -3.10 2.56 .074
Affiliation .84 3.75 2.09 3.56 .096
Achievement 3.52 2.44 2.26 2.72 .124
Order -.47 4.43 1.11 4.34 .127
Nurturance 1.16 2.82 3.30 2.49 .232
Childbearing Desires 17.50 3.18 18.11 2.95 -.61 3.50
.350
Child-Number Desires 2.61 .95 2.67 .97 -.06 .99
.465
Child-Timing Desires 2.70 1.69 2.55 1.59 .15 1.63
.510
Positive CB Motivation 78.99 11.55 81.78 10.90 -2.79 12.94
.336
Negative CB Motivation 49.82 7.36 46.14 7.34 3.68 9.81
.110
Responsibility, Whether 2.86 .45 2.82 .46 .03 .57
.221
Responsibility, When 2.80 .61 2.77 .63 .03 .72
.318
Employment Status 1.96 .12 1.69 .35 .28 .38
-.064
Work Satisfaction 3.89 .63 3.77 .69 .12 .88
.114
Communication, Whether 1.55 .58 1.51 .54 .04 .61
.421
Communication, When 1.44 .47 1.43 .44 .00 .47
.469
Agreement, Whether 1.41 .48 1.39 .46 .02 .50
.421
Agreement, When 1.36 .39 1.34 .38 .02 .42
.396
Marital Problems 3.45 .88 3.52 .83 -.08 .60
.756
21
the motivational antecedents to childbearing. Male parity changers had a mean initial
childbearing desires score of 17.50 compared to male parity non-changers
Table 2. Means and Standard Deviations for Husbands and Wives and for Husband-Wife
Differences and Husband/Wife Correlations for All Predictor Variables at Baseline
Among Parity Non-Changers (N = 149).
Variable Husband Wife Husb-Wife
Husb/Wife
Mean S.D. Mean S.D. Mean S.D.
Correl.
Age 32.10 5.81 30.13 4.65 1.96 3.89
.745
Education 16.96 2.43 15.89 2.35 1.07 2.40
.498
Protestant .49 .50 .44 .50 .05 .68.0
85
Roman Catholic .30 .46 .31 .46 -.01 .60
.141
Other Religion .11 .31 .13 .33 -.02 .36
.387
No Religion .11 .31 .13 .33 -.02 .43
.127
Religiosity 4.08 1.77 4.57 1.77 -.49 1.51
.639
Number of Marriages 1.10 .32 1.11 .33 -.01 .41
.213
Marital Duration 4.18 3.25 4.20 3.21 -.02 .38
.993
Gravidity .57 .81 .61 .87 -.04 .46.8
50
Parity .42 .49
Autonomy -.39 2.86 -2.71 3.10 .090
Affiliation .24 3.68 1.79 3.82 .056
Achievement 3.59 2.41 2.07 2.95 .090
Order -.32 4.79 1.12 4.53 .094
Nurturance 1.17 2.91 2.73 2.82 .308
Childbearing Desires 14.68 4.96 14.46 5.48 .21 4.65
.607
Child-Number Desires 2.25 1.03 2.19 1.11 .06 1.02
.549
22
Child-Timing Desires 3.91 2.41 3.98 2.41 -.07 2.48
.471
Positive CB Motivation 73.47 14.24 73.76 15.27 -.29 14.88
.493
Negative CB Motivation 50.91 8.95 49.32 9.42 1.59 12.15
.125
Responsibility, Whether 2.83 .53 2.72 .60 .11 .62
.415
Responsibility, When 2.77 .65 2.76 .63 .01 .75
.320
Employment Status 1.95 .17 1.80 .31 .15 .34
.059
Work Satisfaction 3.87 .61 3.76 .78 .11 1.01
-.050
Communication, Whether 1.78 .73 1.81 .75 -.03 .75
.489
Communication, When 1.69 .64 1.69 .66 .01 .60
.583
Agreement, Whether 1.66 .68 1.73 .77 -.07 .61
.650
Agreement, When 1.52 .60 1.57 .63 -.05 .53
.629
Marital Problems 3.64 1.09 3.85 1.19 -.21 .81
.750
for both sexes on child-number and child-timing desires. Both sexes showed large
differences across groups on Positive Childbearing Motivation but only females showed a
significant difference (p <.001) on Negative Childbearing Motivation. Additionally, both
sexes indicated significantly worse communication about if and when to have a child in
the non-changers group. One other noteworthy difference was that the female changers
had stronger initial childbearing desires than their husbands (p = .019), whereas female
non-changers actually had weaker desires (although p = .54) than their husbands.
Tables 3 and 4 present descriptive statistics by sex for the two groups of actual
childbearing desires and related difference variables at each of the five interviews. Again
we observe important group differences. For example, husband and wife childbearing
desires dropped by almost 50% over the five years in the parity changers group but by
only about 10% in the non-changers group. Additionally, the latter group started off
23
considerably lower than the parity changers and ended up considerably higher.
Comparing the two groups on Absolute SARA reveals that non-changers had much
higher baseline disagreement in childbearing desires (p <.001). Further, whereas
disagreement is essentially flat across the five year study period for non-changers, it
increases by almost 50% across that same time among parity changers.
It should be noted that the loss to follow-up is greater in the non-changers group,
40.9% versus 12.9%. The primary reason for loss to follow-up in the entire sample was
marital dissolution. Even though one or both members of separated or divorced couples
continued to participate in the study, they could not be included in couple analyses. This
reason for loss was especially important in the non-changers group, both because of their
initial higher level of marital problems (see Tables 1 and 2) and because the absence of
children in more than one-half of the group probably made separation less onerous for
them. Although this difference between groups may have had some impact on our
findings, it does not appear to have had any obvious affect on Absolute SARA. The
mean
24
Table 3. Means and Standard Deviations of Husband and Wife Actual and Perceived Childbearing Desires
and Related Differences Variables in the Parity Changers Group at Each of Five Interviews.
Baseline Follow-up 1 Follow-up 2 Follow-up 3 Follow-up 4
Variable Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D.
(N = 240) (N = 240) (N = 236) (N = 221) (N = 209)
Ha 17.50 3.18 15.55 4.94 13.87 5.93 11.06 6.57 9.10 6.36
Wa 18.11 2.95 15.90 5.08 14.32 6.15 11.66 6.61 9.61 6.23
Ha-Wa -.61 3.50 -.35 3.80 -.45 3.92 -.60 4.73 -.51 4.83
|Ha-Wa| 2.23 2.76 2.53 2.85 2.59 2.97 3.13 3.59 3.31 3.54
Table 4. Means and Standard Deviations of Husband and Wife Actual and Perceived Childbearing Desires
and Related Differences Variables in the Parity Non-Changers Group at Each of Five Interviews.
Baseline Follow-up 1 Follow-up 2 Follow-up 3 Follow-up 4
Variable Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D.
(N = 149) (N = 148) (N = 132) (N = 102) (N = 88)
Ha 14.68 4.96 14.23 5.26 14.07 5.40 13.76 5.41 13.02 5.90
Wa 14.46 5.48 14.09 5.70 13.85 6.12 13.41 5.97 12.80 6.46
Ha-Wa .21 4.65 .14 4.85 .22 4.83 .35 5.43 .21 5.06
|Ha-Wa| 3.41 3.16 3.54 3.30 3.52 3.31 3.90 3.77 3.29 3.83
score on the latter for the non-changers was slightly higher at the third follow-up and
slightly lower at the fourth follow-up when compared with the means for the first three
measurements.
A separate analysis of variance of Absolute SARA by time for the changers group
showed a linear increase in disagreement from the first to the fifth interview and both a
linear and a significant deviation from linear increase as parity rose from zero to three (p
<.001). In the course of the HLM analysis we also discovered that being pregnant at the
time of the interview was a time-varying variable that increased disagreement.
Accordingly, the final Level 1 HLM model of the absolute SARA variable in the
changers group contains terms for the intercept, being currently pregnant, parity, and
parity-squared:
(Absolute SARA)it = π0i + π1i (Currently Pregnant)it
+π2i (Parity)it + π3i (Parity)2it + eit
The Level 2 HLM model includes predictors for the Level 1 parameters as shown in
Table 5. The residual (error) variances for the slope of being currently pregnant (π1i) and
for the slope of parity-squared (π3i) were found to be small and are, therefore, set to zero
in the model. Thus the only covariance components estimated are for the intercept (π0i)
and parity (π2i). Both of these parameters exhibit substantial residual parameter variance.
25
There are eleven predictors of the intercept of Absolute SARA shown in Table 5,
three of which have been added because of their presence in the parity slope model in
order to avoid distorting that slope.2 As Absolute SARA increases, the amount of couple
disagreement increases. With this in mind, the following intercept model variables
predict greater initial disagreement: the husband being relatively older than the wife, a
high Achievement score by the wife, high Negative Childbearing Motivation scores by
both husband and wife, the husband believing himself to have more responsibility for
deciding whether to have a(nother) child than does the wife, and both husband and wife
feeling that communication about when to have a(nother) child has been relatively poor.
The following intercept model variables predict greater initial agreement: husband and
wife having more education between them, a high Affiliation score by the wife, a high
Nurturance score by the husband, and high Positive Childbearing Motivation scores by
both husband and wife.
2The marital problems variable present in the currently pregnant slope model does not need to be added
because of the dichotomous nature of the currently pregnant variable.
26
There is one predictor of the Currently Pregnant slope of Absolute SARA: when
the husband perceives relatively greater marital problems than the wife, there tends to be
greater couple disagreement on childbearing desires during pregnancy. Given the small
mean difference between spouses in marital problems (see Table 1), the Currently
Table 5. Hierarchical Linear Model of Absolute SARA in the Parity Changers Group (N = 240).
Model
Predictor Variable Gamma S.E. T P
Intercept 2.596 1.257 2.066 0.039
Age – Difference 0.082 0.034 2.381 0.017
Education – Mean -0.154 0.051 -3.001 0.003
Roman Catholic – Diff. -0.408 0.268 -1.520 0.128
Achievement – Wife 0.074 0.040 1.837 0.066
Affiliation – Wife -0.058 0.032 -1.828 0.067
Nurturance – Husb. -0.134 0.041 -3.260 0.001
Positive CB Motivation – Mean -0.087 0.017 -5.130 <0.001
Negative CB Motivation – Mean 0.052 0.019 2.786 0.006
Responsibility, Whether – Diff. 0.701 0.193 3.629 0.001
Responsibility, When – Mean -0.418 0.275 -1.520 0.128
Communication, When – Mean 0.662 0.279 2.374 0.018
Currently Pregnant Slope 0.821 0.241 3.405 0.001
Marital Problems – Diff. 0.737 0.373 1.976 0.048
Parity Slope 0.227 1.141 0.199 0.843
Roman Catholic – Diff. 0.620 0.230 2.699 0.007
Positive CB Motivation – Mean 0.036 0.014 2.543 0.011
Responsibility, When – Mean 0.647 0.245 2.641 0.009
Parity Squared Slope 0.290 0.136 2.130 0.033
Note: Abbreviations include Diff. = Difference, Husb. = Husband, and CB = Childbearing
27
Pregnant slope is positive in most cases, meaning that couple disagreement generally
increases during pregnancy. Table 5 also indicates that there are three predictors of the
Parity slope of Absolute SARA. When the husband is Roman Catholic and the wife is
not, when husband and wife have greater combined Positive Childbearing Motivation
scores, and when they assign greater combined responsibility to the husband for deciding
when to have a(nother) child, there tends to be greater couple disagreement on
childbearing desires across parity changes. The coefficients for these variables always
lead to a predicted Parity slope that is positive, meaning that couple disagreement
increases with parity. Finally, there are no predictors of the Parity Squared slope of
Absolute SARA, although that slope is significantly different from zero, indicating that
the slope for parity increases as parity increases and, therefore, that couple disagreement
increases at a faster rate with higher parities.
For those couples who did not change parity during the five years of the study, the
Level 1 HLM model is simpler:
(Absolute SARA)it = π0i + π1i (Time)it + eit
The Level 2 HLM model includes predictor for both parameters, as shown in Table 6.
There is substantial residual parameter variance for both parameters, but less for the time
slope. There are twelve predictors of the intercept of Absolute SARA, four of which
have been added because of their presence in the slope. The following intercept model
variables predict greater disagreement: greater combined religiosity on the part of both
husband and wife, high Negative Childbearing Motivation scores by both husband and
wife, greater employment satisfaction for the husband relative to the wife, a husband
believing himself to have more responsibility for deciding when to have a(nother) child,
and both
28
husband and wife feeling that communication about whether to have a(nother) child has
been relatively poor. The following intercept model variables predict greater agreement:
a high Achievement score by the wife, high Positive Childbearing Motivation scores on
the part of both husband and wife, a husband believing himself to have more
Table 6. Hierarchical Linear Model of Absolute SARA in the Parity Non-changers Group (N = 149).
Model
Predictor Variable Gamma S.E. T P
Intercept 7.385 1.613 4.578
<0.001
Education – Diff. -0.065 0.083 -0.776
0.438
Religiosity – Mean 0.246 0.123 2.006
0.044
Achievement – Wife -0.178 0.062 -2.876
0.004
Affiliation – Husb. 0.070 0.058 1.202
0.230
Autonomy – Husb. 0.099 0.075 1.332
0.183
Positive CB Motivation – Mean -0.046 0.015 -3.048
0.003
Negative CB Motivation – Mean 0.087 0.023 3.704
<0.001
Work Satisfaction – Diff. 0.673 0.184 3.653
<0.001
Responsibility, Whether – Husb. -1.609 0.399 -4.031
<0.001
Responsibility, When – Husb. 0.658 0.325 2.021
0.043
Communication, Whether – Mean 1.219 0.299 4.007
<0.001
Marital Problems – Diff -0.464 0.236 -1.967
0.049
Time Slope -0.649 0.269 -2.410
0.016
Education – Diff. 0.060 0.026 2.291
0.022
Religiosity – Mean -0.101 0.043 -2.352
0.019
Affiliation – Husb. -0.038 0.018 -2.038
0.041
Autonomy – Husb. -0.044 0.023 -1.928
0.053
Note: Abbreviations include Diff. = Difference, Husb. = Husband, and CB = Childbearing
29
responsibility for deciding whether to have a(nother) child, and a husband perceiving
relatively greater marital problems than his wife.
As Table 6 shows, there are four predictors of the time slope of Absolute SARA.
When the husband has more education than his wife, disagreement tends to increase over
time. On the other hand, when husband and wife have greater combined religiosity, the
husband has a high score on Affiliation, and the husband has a high score on Autonomy,
then agreement tends to increase over time.
Discussion
The theoretical approach that guides the analyses reported here rests on two
models of human fertility motivation, one in which individual motivation is understood
as progressing through a four step traits-desires-intentions-behavior sequence and one in
which each member of the dyad apprehends the motivational sequence of his or her
partner. The four processes that we have postulated to affect the respondent’s own
motivation and the perception of partner’s motivation are broadly conceptualized in the
sense that each has been addressed much more specifically by a variety of social-
psychological theories. For this reason, our approach may best be characterized as a
meta-theory, that is, as a general theory about how more specific theories may be
organized in their application to the motivational interaction of couples. Mindful of this,
we call our approach a theoretical framework. Because our theoretical framework is
neutral with respect to the exact nature of the four interaction processes, any of the many
extant social-psychological theories about how these processes work may be tested
within the framework and methodology described here. An important goal of the
analyses discussed below is to suggest directions that such testing might take in the
future.
In the interpretation of our results, we will emphasize how two of the four
motivation-related processes of couples –acceptance and accommodation- affect one
Table 6. Hierarchical Linear Model of Absolute SARA in the Parity Non-changers Group (N = 149).
Model
Predictor Variable Gamma S.E. T P
Intercept 7.385 1.613 4.578
<0.001
Education – Diff. -0.065 0.083 -0.776
0.438
Religiosity – Mean 0.246 0.123 2.006
0.044
Achievement – Wife -0.178 0.062 -2.876
0.004
Affiliation – Husb. 0.070 0.058 1.202
0.230
Autonomy – Husb. 0.099 0.075 1.332
0.183
Positive CB Motivation – Mean -0.046 0.015 -3.048
0.003
Negative CB Motivation – Mean 0.087 0.023 3.704
<0.001
Work Satisfaction – Diff. 0.673 0.184 3.653
<0.001
Responsibility, Whether – Husb. -1.609 0.399 -4.031
<0.001
Responsibility, When – Husb. 0.658 0.325 2.021
0.043
Communication, Whether – Mean 1.219 0.299 4.007
<0.001
Marital Problems – Diff -0.464 0.236 -1.967
0.049
Time Slope -0.649 0.269 -2.410
0.016
Education – Diff. 0.060 0.026 2.291
0.022
Religiosity – Mean -0.101 0.043 -2.352
0.019
Affiliation – Husb. -0.038 0.018 -2.038
0.041
Autonomy – Husb. -0.044 0.023 -1.928
0.053
Note: Abbreviations include Diff. = Difference, Husb. = Husband, and CB = Childbearing
30
correspondence state: the degree of actual agreement between partners. We leave for a
separate analysis (Miller & Pasta, 2004) a consideration of how the remaining two
processes –apprehension and attribution- affect the remaining two correspondence states:
accuracy of perception and perceived agreement. That does not mean that apprehension
and attribution play no role in the changes in couple agreement that occur in response to
reproductive events or simply with the passage of time. Clearly they do and we will
point this out when such an interpretation seems warranted. What it does mean is that we
consider acceptance and accommodation to be the ultimate determinants of changes in
couple disagreement.
Before interpreting our results, it is important to emphasize two constructs that are
implicit in our theoretical framework and that flow from our research design. First,
because the period of study extends through five of the most important years for our
married couples to make decisions about whether and when to have children and how
many to have, we are actually examining change in couple’s motivational disagreement
across a major portion of the reproductive life course. It is for this reason that both time
and reproductive events are so central to understanding our findings. It is also for this
reason that the division of our sample into parity-changer and parity-non-changer groups
is so instructive. Second, when we examine the childbearing desires of two members of a
couple across the reproductive life course, the changes that take place in one or both of
them occur not just in response to reproductive events or the passage of time but also as a
result of their own preexisting traits. One way of expressing this idea is to say that
couples may have a latent disagreement that only becomes manifest in response to a
fertility event or with the passage of time. Apart from measurement error, this
phenomenon has a number of potential causes. The two members of a couple may have
differential satiation of their motivation (Miller & Pasta, 1995) in response to the birth of
a child. Or, they may agree in their childbearing desires but not in their child-number
desires, with the result that childbearing disagreement only appears after the first or
31
second child is born. Alternatively, there may be some non-motivational characteristic of
one partner that affects how that individual or the couple itself responds to events or the
passage of time. Whatever the cause, this kind of latent disagreement can interfere with
acceptance and accommodation between partners across the life course.
Turning to the results, our analyses show that among those who changed parity
during the five-year study period, disagreement varied across two reproductive variables,
being currently pregnant and having a parity increase. On the other hand, among those
who did not change parity, disagreement varied across time. Thus both of these groups
manifested change across the life course. What is so interesting about these two groups is
how different they are motivationally, both with respect to their baseline state and how
they change during the five-year follow-up period.
In the parity changers group (see Table 3), both husbands and wives show a
relatively rapid decline in childbearing desires across the five years of study, with wives
always having on average stronger desires than husbands by a relatively constant amount.
Combining the two sexes, desires decline from a very high mean of 17.80 to one of 9.35
or by almost one-half (47.5%). While this decline is occurring, the absolute difference
between husbands’ and wives’ desires is increasing steadily from 2.23 to 3.31 or by
almost one-half (48.4%). In contrast, in the parity non-changers group (see Table 4),
both husbands and wives show a relatively slow decline in childbearing desires across the
five years of the study, with husbands always having stronger desires by a relatively
constant amount. Combining the two sexes, desires decline from a moderately high mean
of 14.57 to one of 12.91 or by only 11.4%. While this slight decline is occurring, the
absolute difference between husbands’ and wives’ desires is not changing.
These differences suggest that among couples who have children during the five
year study period, it is typically the wife who is leading the way motivationally. These
couples' primary decisions are more likely to be when to have (additional) children and
then how many to have. As they have children, their desires decrease steadily at an
32
annual rate of about 12% (calculations not shown) and this decline gives no evidence of
slowing after five years of follow-up. In the course of these changes, their disagreement
increases and begins to approach the level of disagreement in the parity non-changers.
The increased disagreement that occurs with pregnancy and then follows with each birth
probably reflects, among other things, differences within couples over the desired timing
of additional children and over the desired number of children to have. Among
couples who do not have children during the five years of study, it is the husband who
tends to be more motivated. These couples’ primary decisions appear to be whether to
have (additional) children and then when to have them. As times passes, their desires
decrease only slightly and they maintain high levels of disagreement, which probably
reflects continuing disagreement over whether to have children and when to have them.
These comments set the stage for consideration of the HLM results. In discussing
these, we concentrate on the slope rather than the intercept components of the models.
We do this both because of space considerations and because our greatest interest lies in
the dynamic changes that the slopes represent. There are, however, two important
features of the intercept components of these models that deserve emphasis. First, the
relationships between predictor variables and a given slope are adjusted by the variables
present in the intercept component of each model. Second, the variables that predict the
intercepts represent "good bets" for additional variables that might predict the slopes,
given a little more power or a slightly different sample.
Before discussing the results, we should point out that the two groups we analyze
here are each rather complex in themselves. The parity changer group is a mixture of
those who had planned pregnancies and those who had unplanned pregnancies that were
not aborted (there were few abortions in this sample). As reported elsewhere (Miller,
1994), unplanned pregnancies that are not aborted and that occur in the context of
marriage generally became fully wanted shortly after the baby is born. This suggests that
33
the admixture of unplanned pregnancies probably had the greatest impact for the parity
changer group on the currently pregnant findings.
The parity non-changer group, on the other hand, is a mixture of those who have
avoided (further) childbearing over five years by choice, those who did so because of
subfecundity, and those who agreed to delay (further) childbearing during the five years
of study. By examining our respondents’ answers to final follow-up questions about how
many (more) children they wanted and how they perceived their own fecundity (results
not shown), we estimate that the first two of these three categories together probably total
about a quarter of the non-changers group. Both of these categories can be expected to
have high couple agreement on childbearing desires, although the first with low mean
couple desires and the second with high mean couple desires. Given the high mean
disagreement level in the non-changers group as a whole and given that only a few
couples without disagreement delay beyond five years, most of the remaining three-
quarters are probably couples where disagreement is playing an important role in their
delay.
In the parity changers group, the predictors of increase in disagreement across a
reproductive event all suggest the emerging influence of factors that were latent before
the event and that make acceptance and accommodation within the dyad more difficult.
The greater disagreement demonstrated by couples where the husband is Roman Catholic
and the wife is not seems to represent a situation where religious background helps
sustain the husband's childbearing desires more than the wife’s as parity increases. What
about when the wife is Roman Catholic and the husband is not? In this case, because
wives tend to lead the way in this group, the religious difference reinforces rather than
conflicts with this pattern. The greater disagreement that occurs when both spouses have
high positive childbearing motivation probably occurs because combined high motivation
promotes acceptance and accommodation and leads couples to have children sooner and
to have more of them. However, as these fertility events occur, there is a greater risk than
34
in couples with medium or low motivation that one spouse will reach a limit and
experience a satiation of desires.
Disagreement in childbearing desires increases across parity when both spouses
initially attribute more responsibility for deciding when to have children to the husband.
This suggests that, after a birth whose timing was strongly influenced by the husband,
conflict about the occurrence or timing of an additional birth may emerge. This response
is not surprising given the idea we have developed that it is the wives who typically lead
the way motivationally in this group. It is also not surprising given our previous research
(Miller & Pasta, 1994) which showed that wives placed much more emphasis on their
own desires than on their husbands' during the formation of child-timing intentions,
whereas husbands placed about equal emphasis. Finally, we note the significance of the
parity squared slope. Although parameter variance was too small to model predictors of
this slope, its statistical significance in the model indicates that the increase in couple
disagreement itself increased with each additional parity change during the five years of
follow-up.
The fact that being currently pregnant increases spouse disagreement about
childbearing desires suggests that during pregnancy couples focus more of their attention
on having a child and what the implications are for both members of the couple. This
“thinking through” process then promotes the recognition of individual differences and
thus increases each partner’s apprehension of the other’s desires. In the companion paper
to this one (Miller & Pasta, 2004), our analysis of the predictors of attribution and
apprehension supports this idea that increased accuracy in the perception of partner’s
childbearing desires occurs when couples experience fertility events. It also supports the
conclusion that increased accuracy is associated with a reduction of self bias in
perception. By decreasing perceived agreement, the reduction of self bias in turn tends to
lessen acceptance and accommodation, which is then likely to be sustained after birth. It
35
seems probable that this overall pattern would be especially true if the pregnancy were
unplanned.
The increase in disagreement during pregnancy is greater when the husband
perceives more marital problems than does the wife, suggesting that marital problems
reduce the husband’s acceptance and accommodation when he is confronted with the
impending birth of a child. When the wife perceives more problems, on the other hand,
her typically greater motivation to bear a child may sufficiently overcome her concern
about the marriage to allow her to remain accepting and accommodating regarding any
additional childbearing. Again this pattern may be especially applicable in the case of
unplanned pregnancies.
Whereas in the parity changers group the predictors of increased disagreement are
related to the occurrence of fertility events, in the parity non-changers group the
predictors are related to the non-occurrence of such events. Thus the important predictors
are those that influence disagreement simply with the passage of time as the couple
considers whether to have a(nother) child. The effect of educational differences is
perhaps most readily understood in terms of the wife. We already know that more
education tends to decrease the positive motivation for childbearing among the women of
this sample (Miller, 1992). This suggests that an educational differential favoring the
wife reinforces the basic motivational pattern in this group, promoting her acceptance of
and accommodation to childbearing delay. On the other hand, an educational differential
favoring the husband may well mean she has diminished career options, therefore
reducing her acceptance and accommodation.
The greater agreement that occurs when the combined religiosity of the couple is
high may come because the prosocial and promoral aspects of high religiosity make each
spouse more willing to be accepting and accommodating of the other's desires, whether
they are inclined not to have a(nother) child, delaying its birth, or are experiencing
difficulty in conceiving. The fact that religiosity works in this way only in the parity
36
non-changers group may reflect the important support that religion can provide in making
a counter-normative decision.
The fact that high scores by a husband on the personality traits of affiliation and
autonomy promote greater agreement over time suggests two factors that may allow him
to be more accepting of his wife’s weaker childbearing desires and more accommodating
to his wife’s perceptions of his desires as weak. An affiliative husband may be better
able to meet his social needs other than through (further) childbearing. Further, the
results of the companion paper to this one (Miller & Pasta, 2004) indicate that he may
also be more biased toward self in the perception of his wife’s desires and therefore less
aware of their potential for disagreement. An autonomous husband, on the other hand,
may be more comfortable making the kind of counter-normative decision (no children or
a single child) that this group seems to be making.
In summary, our findings support the central role of motivational disagreement in
fertility decision-making. The results with the group of respondents who made no parity
changes during the five years of study confirm the findings of a growing number of
studies which suggest that motivational disagreement can play an important role in the
delay of fertility and, ultimately, in the amount of childbearing experienced by couples.
On the other hand, the results with the group who did make parity changes demonstrate
how motivational disagreement within couples increases progressively with each fertility
event that they experience, ultimately acting –along with a decrease in their mean
motivational level- as a kind of brake on further fertility. Thus couple disagreement
about childbearing desires, which can be viewed as a specific example of reproductive
conflict of interest between two mated partners, has been found here to be a key part of
the dynamic interactive process that determines the number and timing of children born
to a couple.
37
References
Bagozzi, R.P., & Van Loo, M.F. (1987). Individual and couple tastes for children:
Theoretical, methodological, and empirical issues. Journal of Economic
Psychology, 8, 191-214.
Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical linear models: Applications and
data analysis methods. Newbury Park, CA: Sage.
Bryk, A.S., Raudenbush, S., Seltzer, M., & Congdon, R. (1988). An Introduction to
HLM: Computer Program and User's Guide (2nd Ed.). Chicago: University of
Chicago Department of Education.
Eisenberg, N, & Miller, P. (1987). The relation of empathy to prosocial and related
behaviors. Psychological Bulletin, 101, 91-119.
Jackson, D.N. (1984). Personality Research Form Manual. Port Huron, MI: Research
Psychologist Press.
Miller, W.B. (1992). Personality traits and developmental experiences as antecedents of
childbearing motivation. Demography, 29, 265-285.
Miller, W.B. (1994). Childbearing motivations, desires, and intentions: A theoretical
framework. Genetic, Social, and General Psychology Monograms, 120, 223-258.
Miller, W.B. (1995). Childbearing motivation and its measurement. Journal of
Biosocial Science, 27, 473-487.
Miller, W.B., & Pasta, D.J. (1993). Motivational and non-motivational determinants of
child-number desires. Population and Environment, 15, 113-138.
38
Miller, W.B., & Pasta, D.J. (1994). The psychology of child timing: A measurement
instrument and a model. Journal of Applied Social Psychology, 24, 218-250.
Miller, W.B., & Pasta, D.J. (1995a). Behavioral intentions: Which ones predict fertility
behavior in married couples? Journal of Applied Social Psychology, 25, 530-555.
Miller, W.B., & Pasta, D.J. (1995b). How does childbearing affect fertility motivations
and desires? Social Biology, 42, 185-198.
Miller, W.B., & Pasta, D.J. (1996a). The relative influence of husbands and wives on
the choice of oral contraception, a diaphragm, and condoms. Journal of Applied
Social Psychology, 26, 1749-1774.
Miller, W.B., & Pasta, D.J. (1996b). Couple disagreement: Effects on the formation and
implementation of fertility decisions. Personal Relationships, 3, 307-336.
Miller, W.B., & Pasta, D.J. (2004). Bias toward self in the perception of spouse’s
childbearing desires. Submitted for publication.
Miller, W.B., Severy, L.J., & Pasta, D.J. (2004). A framework for modeling fertility
motivation in couples. Population Studies, 58, 193-205.
Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind?
Behavioral and Brain Sciences, 1, 515-526.
Thomson, E. (1983). Individual and couple utility of children. Demography, 20, 507-
518.
Thomson, E. (1997). Couple childbearing desires, intentions, and births. Demography,
34, 343-354.
Thomson, E., McDonald, E., & Bumpass, L.L. (1990). Fertility desires and fertility:
Hers, his, and theirs. Demography, 27, 579-588.
39
