Advances in Life Course Research 23 (2015) 14–28
Contents lists available at ScienceDirect
Advances in Life Course Research
journal homepage: www.elsevier.com/locate/alcr
Fertility intentions and outcomes
Implementing the Theory of Planned Behavior with
graphical models
Letizia Mencarini a,b,*, Daniele Vignoli c, Anna Gottard c
a
b
c
University of Turin, Italy
Collegio Carlo Alberto, Italy
University of Florence, Italy
A R T I C L E I N F O
A B S T R A C T
Article history:
Received 9 February 2014
Received in revised form 2 October 2014
Accepted 15 December 2014
This paper studies fertility intentions and their outcomes, analyzing the complete path
leading to fertility behavior according to the social psychological model of Theory Planned
Behavior (TPB). We move beyond existing research using graphical models to have a
precise understanding, and a formal description, of the developmental fertility decisionmaking process. Our findings yield new results for the Italian case which are empirically
robust and theoretically coherent, adding important insights to the effectiveness of the
TPB for fertility research. In line with TPB, all intentions’ primary antecedents are found to
be determinants of the level of fertility intentions, but do not affect fertility outcomes,
being pre-filtered by fertility intentions. Nevertheless, in contrast with TPB, background
factors are not fully mediated by intentions’ primary antecedents, influencing directly
fertility intentions and even fertility behaviors.
ß 2014 Elsevier Ltd. All rights reserved.
Keywords:
Fertility intentions
Theory of Planned Behavior
Graphical models
Italy
1. Introduction
The study of fertility intentions has become central in
the discussion of fertility rates in developed countries,
under the realistic assumption that, in an almost perfect
contraceptive regime, having a child is a result of a
reasoned, although imperfect, decision. Whereas fertility
intentions have been a central theme in demographic
research for some time, it has received renewed attention
in recent years since intentions are now frequently
analyzed in the framework of Theory Planned Behavior
(hereafter, TPB), a general psychological theory concerning
the link between attitudes and behavior (Ajzen, 1991,
2005; Ajzen & Fishbein, 1980; Fishbein & Ajzen, 2010).
Billari, Philipov, and Testa (2009), as well as Ajzen and
* Corresponding author at: University of Turin, Italy.
E-mail address: letizia.mencarini@unito.it (L. Mencarini).
http://dx.doi.org/10.1016/j.alcr.2014.12.004
1040-2608/ß 2014 Elsevier Ltd. All rights reserved.
Klobas (2013), specifically discuss the possible application
of TPB in the fertility domain. Fertility outcomes, according
to TPB, are seen as depending directly on fertility
intentions, which in turn depend directly on attitudes
(related to the perceived benefits and/or costs of reproduction), subjective norms (related to the social approval
of behavior from relevant others), and perceived behavioral control. Possible constraints can further intervene
from the time the fertility intention was formed and the
subsequent behavior (such as a disruption of the couple’s
relationship or changes in individuals’ health conditions or
job status). This multi-factor paradigm is expected to
depend on several background factors as well (such as
socioeconomic and demographic factors). Whether TPB is a
valid framework for analyzing human fertility is, however,
a hotly debated issue (Ajzen, 2011; Barber, 2011; Klobas,
2011; Liefbroer, 2011; Miller, 2011; Morgan & Bachrach,
2011; Philipov, 2011) and, apart from the theoretical
debate, there are only few attempts to test the TPB
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
framework in its full complexity (e.g., Dommermuth,
Klobas, & Lappegård, 2013; Kuhnt & Trappe, 2013).
In this study, we aim to analyze the complete path
leading to fertility behavior, within the explanatory
framework of TPB and considering the most common
background variables (i.e., determinants) of fertility. We
move beyond existing research developing a unified
empirical strategy to test the validity of TPB. Through
graphical models, we provide a precise understanding, and
a formal description, of the developmental fertility
decision-making process by studying the dependencies
among all the factors involved in the TPB, on the basis of
their joint distribution. In our application, variables are
partitioned into a sequence of blocks: we can distinguish
between pure response variables (in the last block), pure
background variables (in the first block), and intermediate
variables, which are responses for variables in previous
blocks and explanatory for the subsequent variables. The
partial ordering among the variables into blocks is fully
derived by the TPB. The method, while not materially
different from more conventional ones, does provide a
useful conceptual fit to the TPB and facilitate its evaluation.
Our study focuses on Italy, a country for which the use
of TPB in fertility research is a novelty. We rely on a specific
module questionnaire, within the 2003 Italian GGS survey,
that was designed to collect the relevant dimensions of the
TPB (Vikat et al., 2007). Then, using the 2007 follow-up of
that survey, we look at the subsequent fertility behavior,
thus completing the study of the whole process leading up
to the decision to have a child.
2. Theoretical background on the study of fertility
intentions and realizations
2.1. The Theory of Planned Behavior as a theoretical
framework for the fertility decision-making process
According to the TPB, which is an extension of the
earlier Theory of Reasoned Action (Fishbein & Ajzen, 1975,
2010), intentions are the immediate antecedents of
corresponding behavior. This hypothesis is supported by
several systematic reviews of the empirical literature, and
strong intention–behavior correlations are also observed
in the fertility domain (Ajzen, 2010; Billari et al., 2009). As
Ajzen and Klobas (2013) advocate, however, a concern in
applying the TPB in the study of human fertility is defining
an appropriate behavioral criterion. In fertility research,
having a child, is commonly described as a behavior.
However, strictly speaking, a child birth is an outcome of
specific antecedent behaviors (e.g., having sex, not using a
contraceptive, using artificial reproductive technology,
etc.) that (may) result in a pregnancy. From this perspective, having a child is an outcome or behavioral goal, rather
than a behavior, which might result in attainment of the
goal. Central to this discussion is the implicit or explicit
assumption that, at least in developed countries with
readily available contraception, having a child is the result
of a reasoned decision (e.g., Goldstein, Lutz, & Testa, 2004;
Mills, Mencarini, Tanturri, & Begall, 2008; Testa & Grilli,
2006). In almost perfect contraceptive regimes, in fact, the
difference between fertility as a behavior and fertility as a
15
behavioral goal is as narrow as it can be. Of course,
individuals have greater control over performance of a
behavior than they have over attaining a goal the behavior
is intended to produce.
Ajzen and Klobas (2013) describe how the TPB can be
used to model fertility decision-making process: when
people formulate their intentions to have a(nother) child,
they rely on three conceptually distinct, but interrelated,
primary antecedents of fertility intentions: attitudes,
subjective norms, and perceived behavioral control. These
antecedents represent the most important predictors of
fertility intentions.
The ‘‘attitudes toward the behavior’’, which can be
favorable or unfavorable, are ‘‘readily accessible or salient
beliefs about the likely consequences of a contemplated
course of action’’ (Ajzen, 2010). In the case of fertility
decision-making, individuals would be expected to reflect
on their attitudes about having a child before forming their
fertility intentions. Such attitudes are a person’s internal
evaluation that having a child will have positive or negative
(i.e., desirable or non-desirable) consequences for her/him.
The ‘‘subjective norms’’ are related to the perceived
normative beliefs and expectations of relevant referent
groups or individuals who exert social pressure to perform
or avoid the behavior. In the case of fertility intentions,
individuals would be expected to consider subjective norms
for having a child; i.e., the individual’s perception of the
psychological support of or normative pressure on her/his
fertility behavior from members of her/his close social circle.
Finally, individuals are assumed to take into account
factors that may promote or hinder their ability to perform
the behavior, and these salient control beliefs lead to the
formation of ‘‘perceived behavioral control’’. This refers to
the perceived capability of performing the behavior.
Because many behaviors pose difficulties in execution, it
is useful to consider perceived behavioral control over
having a child in addition to intentions. Like attitudes and
subjective norms, perceptions of behavioral control follow
consistently from readily accessible beliefs about
resources and obstacles that can facilitate or interfere
with the ability to have a child, such as income or wealth,
labor force status, and education (Billari et al., 2009). The
power of each control factor to facilitate, or inhibit,
behavioral performance is expected to contribute to
perceived behavioral control in direct proportion to the
subjective probability that the control factor is present in
each person (Ajzen, 2010).
According to Ajzen and Klobas (2013) fertility intentions are also expected to result in having, or not having, a
child to the extent that people are in fact capable of
attaining their goals, i.e., to the extent that they have actual
control over having a child. Actual behavioral control is
thus hypothesized to moderate the effect of intention on
behavior. Because actual behavioral control is identifiable
with difficulty, perceived control is often used as a proxy
for actual control, under the assumption that perceptions
of control reflect actual control reasonably well.
Demographic research directed toward explaining or
predicting fertility intention within the reasoned action
tradition of the TPB has focused primarily on the intention
to have a child relative to the intention to not have a child
16
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Fig. 1. A schematic presentation of the Theory of Planned Behavior for fertility decision-making.
Adapted from Ajzen and Klobas (2013).
(Billari et al., 2009; Jaccard & Davidson, 1975; Jorgensen &
Adams, 1988) and on the timing of the intentions
(intentions in the short run versus intentions in long
run, Dommermuth, Klobas, & Lappegård, 2011), often
disregarding – mainly because of lack of suitable data – the
complete path leading to fertility behavior.
A schematic representation of the TPB applied to
fertility research is shown in Fig. 1. TPB does not discount
the importance of background factors that can influence
behavior indirectly. They are selected by a content-specific
theory, which can complement the TPB ‘‘by identifying
relevant background factors and thereby extending
understanding of a behavior’s determinants’’ (Ajzen,
2010). Therefore, a number of well-established factors
studied in demographic research, such as age, parity, or
education, are treated as external variables. Under ideal
conditions, in the operationalization of the TPB the
background factors should affect only the primary antecedents of fertility intentions, and should not have a direct
impact on the intentions themselves, or on fertility
outcomes. It is worth noting that it is difficult to
empirically disentangle actual behavioral control and pure
background factors. What, with detailed survey data, can
be identified is the effect of actual enablers and constraints
which intervene between the moment of declaration of
fertility intentions and the time of the child birth.
2.2. Empirical evidence on the determinants of fertility
intentions and realizations
According to the TPB, the distinction between attitudes,
norms, and perceived behavioral control should completely filter the role played by background factors on fertility
intentions, which will in turn determine the subsequent
realization. It is therefore crucial to choose a set of
background factors which have been proven to influence
both fertility intentions and their realization. There is a
plethora of empirical research focusing on the determinants of fertility intentions suggesting that they depend on
several demographic, socioeconomic, and gender-related
factors (e.g., Berrington & Pattaro, 2013; Cavalli & Klobas,
2013; Kapitány & Spéder, 2012; Mills et al., 2008; Neyer,
Lappegard, & Vignoli, 2013; Spéder & Kapitány, 2014;
Thomson, 1997; Thomson, McDonald, & Bumpass, 1990;
Vignoli, Rinesi, & Mussino, 2013). In contrast, the literature
investigating the correlates of the realization of fertility
intentions is scarce, mainly due to a severe lack of
appropriate longitudinal data.
It has been suggested that the factors affecting fertility
intentions are both demographic and socioeconomic. Of
the purely demographic factors, parity and the woman’s
age play crucial roles in the definition of fertility intentions
(Berrington, 2004; Bongaarts, 1992, 2001; Liefbroer, 2009;
Morgan, 1982; Noack & Østby, 1985; Rinesi, Pinnelli, Prati,
Castagnaro, & Iaccarino, 2011; Thomson, 1997). Generally,
the documented findings show that there is an inverse
relationship between fertility intentions and parity (Bühler, 2008; Thomson, 1997). Positive fertility intentions –
i.e., the intentions to have a child – also seem to be less
frequent among older women. The effect of the type of
union has also been widely investigated. Liefbroer (2009)
showed that married women have higher average fertility
intentions than those who do not have a partner or who are
cohabiting. Accordingly, Régnier-Loilier and Vignoli (2011)
have found that, in Italy, cohabiting couples want fewer
children than married couples (however, in France, such
effect has not been found; Régnier-Loilier & Vignoli, 2011;
Toulemon & Testa, 2005).
The role of education was emphasized in a crosscountry study by Heiland, Prskawetz, and Warren
(2008). In many European societies, higher educational
levels were associated with the intention for a greater
number of children. The positive association between
educational level and fertility intentions was also confirmed by a study on France (Toulemon & Testa, 2005) and
by a study on Bulgaria and Hungary (Philipov, Spéder, &
Billari, 2006). Compared to women with less education,
highly educated women displayed better labor market
opportunities and earnings, as well as greater bargaining
power within the couple, which encourages a more equal
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
division of housework and childcare between partners,
and which could in turn facilitate fertility intentions
(Mills et al., 2008).
Among the relatively few studies which have looked at
the transformation of fertility intentions into realization,
demographic factors appear to play a pivotal role. In
particular, woman’s age and parity are crucial (e.g., Noack
& Østby, 2002; Quesnel-Vallée & Morgan, 2003; Rinesi et al.,
2011): postponing motherhood results in having fewer
children than originally planned. Moreover, the greater the
distance between the actual and expected number of
children, the faster the transition toward childbearing in a
short period (Symeonidou, 2000; Thomson et al., 1990). The
type of union is also important. Married couples are more
likely to realize their intention to have a/nother child in the
United States (Quesnel-Vallée & Morgan, 2003; Schoen,
Astone, Kim, Nathanson, & Fields, 1999), Italy, and France
(Régnier-Loilier & Vignoli, 2011). The fertility intentionrealization gap is smallest for highly educated women
(Rinesi et al., 2011; Toulemon & Testa, 2005). Finally, the
effect of gender roles seems to vary in different contexts: in
Greece, less traditional women have greater difficulties than
more traditional women in realizing their positive fertility
intentions (Symeonidou, 2000); while in other contexts,
such as Sweden, the trend is reversed (Thomson, 1997).
3. Data
3.1. Italian Gender and Generation Survey
We use the Italian Gender and Generation Survey and
its corresponding follow-up survey. The Italian variant of
the GGS is a prospective and retrospective survey
conducted by the Italian National Statistical Office (Istat)
in 2003, which is called Family and Social Subjects, or GGSFSS (2003). The follow-up survey, which looked at critical
points in life histories from a gender perspective, was
jointly conducted by Istat and the Ministry of Labour in
2007. GGS-FSS (2003) has a sample of about 24,000
households and 50,000 individuals of all ages (with a nonresponse rate of 17.7%). Out of them, almost 32,000 are
aged 18–64. The follow-up includes about 10,000 interviews with people aged 18–64. In this wave, the overall
non-response rate was 48.6%.
Individuals’ intentions to have a child within the next
three years were surveyed using the following question:
‘‘Do you intend to have a child in the next three years?’’.
The four possible answers were: ‘‘definitely not’’, ‘‘probably not’’, ‘‘probably yes’’, and ‘‘definitely yes’’. Limiting the
question about childbearing intention to a foreseeable
time frame avoided some of the problems associated with
the surveying of intentions. Questions about intentions,
that cover a foreseeable time period, are ‘‘in close temporal
proximity to the prospective behavior’’ (Ajzen & Fishbein,
1973) and are generally considered to be better predictors
of behavior (Billari et al., 2009; Philipov, 2009). They allow
researchers to make inferences based on a person’s current
status about what economic, institutional, and familial
conditions are crucial in her/his decision process to have
a/nother child.
17
Our selected sub-sample consisted of 2871 women,
aged 18–49, married or cohabiting with a partner.1 We
focus on them because the battery of questions related to
TPB was asked only to individuals in co-residing couples.
Overall, linking the two Italian GGS waves allowed us to
assess whether the fertility intentions were subsequently
realized. We covered the period of 2003–2007, during
which the births2 of 368 children were registered.
The association between the birth of a child as the
outcome of interest and prior intentions turned out to be
particularly strong at the extremes: the stronger the
intention to have or not have children, the greater or the
lower the observed proportion of respondents who
realized this intention.3 We found that negative fertility
intentions are a potent predictor of subsequent fertility
behavior. By contrast, positive fertility intentions tended to
overestimate fertility realizations: about 40% of the Italian
respondents who firmly stated the intention to have a child
in the following three years did not achieve their goal.
Incidentally, these results are relatively consistent with
those of a handful of other studies for Italy based on
different data sources (e.g., Rinesi et al., 2011).
3.2. The relevant variables and dimensions for the TPB model
As in all of the Gender and Generation Surveys, the
2003 Italian FSS survey included a battery of questions
needed to implement the TPB (Vikat et al., 2007). Ten items
were utilized to characterize attitudes toward having a
child. Each of these items was introduced by the question:
‘‘If you were to have a/nother child within the next three
years, would it be better or worse in relation to. . .’’. Among
the possible responses were: ‘‘much better’’, ‘‘better’’,
‘‘neither better nor worse’’, ‘‘worse’’, and ‘‘much worse’’.
Subjective norms were measured through three questions.4 The respondents were asked to rate the extent to
which they agree that different groups of people think they
should have a/nother child. All three items were introduced by the following question: ‘‘If you were to have a
child in the next three years, to what extent would the
following persons agree with your choice?’’. Among the
1
In this study, we have deliberately chosen to consider the path
leading to fertility behavior only for women. Reproductive behavior in
advanced societies is most of time a joint decision of the two partners,
therefore conducting a study at the couple level would have been of
substantial interest. In our research framework, we could have estimated
all the graph models specifically for men, or considered men and women
samples pooled together. However, in the first case this would have
produced the double of figures and graphs; in the second case to assess
properly gender issues and looking at couples, we should have tested
several interactions complicating the graphical representation of the
results, and, moreover, this would also have implied considering a
variable that is not embedded within the TPB – i.e., the couple’s
agreement over fertility intentions.
2
These births are live births. Because of lack of specific information we
cannot consider subfecundity (i.e., miscarriages, still births and induced
abortions).
3
See Régnier-Loilier and Vignoli (2011) for a detailed discussion on the
intentions’ predictive strength using the same data.
4
The Italian FSS-GGS omitted one important item included among the
subjective norms according to the GGS guideline (Vikat et al., 2007),
namely ‘‘other relatives’’.
18
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
possible responses were: ‘‘would strongly agree’’, ‘‘would
agree’’, ‘‘would neither agree nor disagree’’, ‘‘would
disagree’’, ‘‘would strongly disagree’’.
Finally, the survey included seven items intended to
capture perceived behavioral control. The following
question was posed for each of the item: ‘‘the decision
about whether to have children can depend on various
situations. How much would your decision about whether
to have a child in the next three years depend on. . .’’.
Among the possible answers were: ‘‘a lot’’, ‘‘quite lot’’, ‘‘a
little’’, and ‘‘not at all’’. In the case of perceived behavioral
control, we reversed the scale, because this made it easier
to show the possible positive effect of the perceived ability
to overcome constraints with a positive coefficient in the
regression model.
We used factor analysis to verify whether the items
available in the used survey acted as valid and reliable
measures of the TPB variables (Billari et al., 2009;
Dommermuth et al., 2011).5 We tested both a three-factor
solution (the proposed factors were attitudes, subjective
norms, and perceived behavioral control) and a four-factor
solution (which allowed for attitudes to fall into two groups),
as was done in Billari et al. (2009) and Dommermuth et al.
(2011). Four factors can be identified (see Table A1 in the
Appendix, where also the list of the used items is reported).
Two can be reasonably interpreted as attitudes factors, one
as a measure of subjective norms, and one as a measure of
perceived behavioral control. The first of these factors is
hence called ‘‘negative attitudes’’, as it represents beliefs
about the costs or personal losses associated with having a
child, while the second is called ‘‘positive attitudes’’ because
it represents beliefs about the benefits of having a child.
4. A chain graph model for implementing the TPB
The TPB itself illustrates a temporal sequence for the
process leading to the decision to have a child. The
existence of such a sequence suggests the possibility of
using the class of graphical models to gain a precise
understanding of the developmental fertility decisionmaking process, as they allow us to study the conditional
independence structure among all of the variables involved
in this process, and depict this structure by a graph.
Graphical models,6 developed as an extension to path
analysis (Wright, 1921), are a class of multivariate models
that are useful for studying, estimating, describing and
visualizing the relationships among an entire set of
variables of interest. Only few applications of this class
of models have been recorded in social sciences (e.g.,
Berrington, Hu, Smith, & Sturgis, 2008; Gottard, 2007;
Mohamed, Diamond, & Smith, 1998). A multivariate model
is graphical whenever its conditional independence
structure can be univocally depicted by a graph. A graph
G = (V, E) consists of two finite sets: a set V for nodes and a
set E collecting edges between nodes. The edges in a graph
5
We used alpha factor analysis of the correlation matrix and then
performed a rotation of the loading matrix through oblimin criteria.
6
See Lauritzen (1996) and Edwards (2000) for a comprehensive
introduction of graphs and graphical models.
can be undirected (lines), or directed (arrows). In graphical
models, nodes represent variables, and the absence of a
connection between two nodes represents a conditional
independence. Graphs are therefore utilized to give a
theoretically rigorous but intuitively easy to understand
representation of the complex relationships among variables, on the basis of their joint distribution. These
relationships are described in terms of conditional
independence, which is the key concept of graphical
models. Two variables, X and Y, are conditionally
independent given a third variable Z, denoted by X E YjZ,
when controlling for Z, X does not provide additional
information on the distribution of Y, and vice versa. A
conditional independence statement is visualized in the
graph by the absence of a connection between two nodes.
In this study we use chain graph models, which are the
most appropriate to empirically implement the temporal
sequence of the fertility decision-making process. They are
in fact graphical models associated with a chain graph
(Lauritzen & Wermuth, 1989): i.e., a graph with both
directed and undirected edges. These graphs, also called
block-recursive graphs, are particularly useful when, as in
our case of the fertility decision-making process, variables
admit a partial ordering on the basis of subject matter
considerations, as hypothesized by the TPB. Variables are
then partitioned into blocks. Those variables belonging to a
same block are considered to be of equal standing, while
those belonging to different blocks can be joined by
arrows, representing a directional association. Consequently, it is possible to distinguish between pure response
variables (in the last block), pure explanatory/background
variables (in the first blocks), and intermediate variables,
which are responses for variables in previous blocks and
explanatory for the subsequent variables. Here we adopt
the Lauritzen–Wermuth–Frydenberg (LWF) class of Markov properties, which establish how conditional independencies can be deduced by a graph (see Lauritzen and
Wermuth (1989), Frydenberg (1990), and Borgoni, Berrington, and Smith (2012), for a detailed presentation on
how read conditional independence in a LWF chain graph).
To better understand how conditional independence
can be read from a chain graph, it is helpful to look at Fig. 2,
which shows a chain graph consisting of five blocks, as in
our case. The first block (a) on the left collects the two
nodes corresponding to the two pure explanatory variables
X1 and X2. The blocks (b), (c) and (d) contain intermediate
variables: each of them is explanatory for variables in the
block on the right, response for those on the left and
concomitant for those in the same block. Finally, the last
block (e) contains the variables X7 and X8, which are
studied only as responses, given all the other variables. For
example, in Fig. 2, the lack of the edge between nodes 1 and
2 corresponds to the marginal independence statement
X1 E X2, as no previous blocks are present. On the other
hand, the absence of the edge between nodes 3 and
4 indicates that X3 E X4jX1, X2.
In a LWF chain graph model as adopted here, estimates
for the parameters of the joint distribution of the variables
was obtained via maximum likelihood. The multidimensional problem has been simplified by factorizing the joint
distribution according to the block structure of the graph,
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
19
Fig. 2. Example of a chain graph with five blocks.
into a sequence of univariate models, as suggested by Cox
and Wermuth (1996). Each univariate model has been
specified according to the nature of the response, where all
of the variables in the same and in the previous blocks are
considered as explanatory. The edges selection was
achieved using a stepwise procedure which compares
the model for the reduced graph with the complete graph
by means of the Likelihood Ratio test (with the significance
level set at 0.05).
In the setting of the chain graph, the TPB guides us to put
variables into blocks. The sequence of the fertility decisionmaking process is produced by ‘‘background variables’’
(block a), ‘‘perceived behavioral control,’’ ‘‘subjective
norms,’’ ‘‘positive and negative attitudes’’ (block b), ‘‘fertility
intentions’’ (block c), ‘‘actual constraints’’ (block d), and
‘‘fertility outcome’’ (block e); see Table A2.
The background variables of the first block (a) were
selected based on evidence from the literature, as outlined
in Section 2.2, as well as based on the peculiarities of the
Italian context (e.g., Dalla Zuanna, 2001; De Rose, Racioppi,
& Zanatta, 2008; Mencarini & Tanturri, 2006; Rinesi et al.,
2011). They are: number of children, woman’s age,
duration of the couple’s relationship, type of couple,
woman’s and man’s educational levels and employment
situations, gender arrangements between the partners,
religiosity, number of siblings,7 macro-region of residence,
municipality size. The dependence structure among the
background variables has not been studied because their
association is far outside of the scope of the paper.
The variables in the second block (b) are the variables at
the core of the TPB. On the basis of the four latent factors
previously identified (see Section 3.2), we constructed four
derived variables.8 These derived variables are considered
7
The number of siblings is indeed the fertility of interviewer’s parents
and is one of the key determinants of fertility intentions: it has been
shown (see, for instance, Régnier-Loilier, 2006) that generally those who
have some siblings have better memories about their childhood
compared to lonely children, and therefore, on average, they tend to
desire for themselves a higher number of offspring.
8
The derived variables were obtained by summing the original
variables and were then considered as ordinal variables. We preferred
derived variables to factor loadings as they are not latent variables, and
can therefore be more easily inserted into the joint distribution of the
chain graph model, while still giving similar information. See Cox (2008)
for the properties of sum-derived variables.
as intermediate variables, as they are dependent on the
background variables and explanatory for the variables in
the following blocks. As dependent variables, they were
modeled by adopting a cumulative logit model for ordinal
variables.
For the variable in the third block (c), we estimated a
cumulative logit model predicting fertility intentions.9 The
fourth block (d) of ‘‘actual constraints’’ includes the binary
variable for the disruption of the couple’s relationship,
which is studied as a logit model. The inclusion of this
block is in line with the TPB and highlights the importance
of intervening factors between the time when fertility
intentions are declared and the fertility outcome. These
factors tend to be neglected by fertility studies that assume
that the determinants of intentions also influence the
subsequent behavior. Some problems or constraints, which
inhibit the realization, can however arise after the moment
when the intentions are expressed, such as unexpected
conflicts between the partners (Régnier-Loilier & Vignoli,
2011).
Finally, the last block (e) consists of the pure response
variable – i.e., the fertility outcome – whose dependence
on the variables in the previous blocks was estimated by
adopting a logit model.
It should be noted that an important logical element in
the TPB is the proposed effect of actual behavioral control
on perceived behavioral control. Fertility intentions are
expected to be transformed into a behavior to the extent
that people are capable of attaining their goals; i.e., to the
extent they have actual control over having a child.
Background factors and actual behavioral controls are
theoretically not well discernible and, as result, there is a
high likelihood that background factors are seen empirically to have a direct effect on intentions. We believe that
the main difference between the two concepts is that
people are able to assess their ‘‘perceived behavioral
control’’ at the moment they form their intention while
they are not able to foresee ‘‘actual constraints’’ after the
intention, and therefore we have chosen the variables
9
Note that the intentions to a have a child are often considered in the
literature as ‘‘parity-progression intentions’’ (Billari et al., 2009).
However, our study had to use a very small-scale sample, and, as a
consequence, we could not stratify our analysis by parity.
20
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
within blocks accordingly (i.e., considering as actual
constraints – block d – only those factors intervening
between the two waves of the survey).
5. Structure of (in)dependence among the variables of
TPB
Overall, the analysis produced a large and interesting
set of empirical findings, but our specific focus is on
dependency structure among the variables, i.e., what is in/
dependent of/on what. We deliberately abstain to describe
in details the direction of the associations we found among
the variables involved in the TPB and we provide insights
regarding the direction of the effects only when it appears
to be relevant for the discourse (the detailed table of
results is in Table A3 in the Appendix). However, it is worth
nothing that the results from the graphical models are fully
consistent with those of other studies on fertility intentions for the Italian case.
Consistently with the theory, looking at Fig. 3 we can
see that the primary antecedents of fertility intentions
considered in the TPB (PAtt, NAtt, SubN, PBC) are
correlated with one other. Only the ‘‘perceived behavioral control’’ (Pbc) is independent of the index standing
for ‘‘subjective norms’’ (SubN), given the ‘‘positive’’
(PAtt) and ‘‘negative attitudes’’ (NAtt), as well as the
background variables (Fig. 3). We can also discern which
background variables influence the primary antecedents
of fertility intentions, conditionally one to another. The
‘‘perceived behavioral control’’ depends on the number
of children (Nch), the current gender division of
housework in the family (CHD), and the woman’s age
(AgeW). The ‘‘positive attitudes’’ dimension depends on
the degree of religiosity (Rel), the number of children
(NCh), and the duration of the couple’s relationship
(CDur). The ‘‘negative attitudes’’ dimension depends on
the number of children (Nch), the current division of
housework (CHD), and the area of residence (Reg). The
‘‘subjective norms’’ dimension depends on the number of
children (Nch), the number of siblings (Sib), the duration
of the couple’s relationship (CDur), the woman’s age
(AgeW), and the number of children (NCh). Overall, the
only background factor influencing all the primary
antecedents of fertility intentions is the number of
children, while some of the background variables do not
influence any of the primary antecedents of fertility
intentions. This fact can be visualized in Fig. 3 by the fact
that they stand alone: the man’s and the woman’s
employment situations (ME, WE), the woman’s satisfaction regarding the housework division (SHD), the
parents’ residential proximity (PRP), the municipality
size (MunS), the type of couple (CTy).
Fig. 3. Conditional independence graph of background variables (block a) and primary antecedents (block b).
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
21
Fig. 4. Conditional independence graph of background variables (block a), primary antecedents (block b), and fertility intentions (block c). Note: Arrows from
block a (background) to block b (primary antecedents) are not shown for simplicity (they were displayed in Fig. 3).
Corroborating the scheme of TPB, the level of fertility
intentions (Fint) depends on all of the predicted primary
antecedents of fertility intentions: (positive and negative)
attitudes, subjective norms, and perceived behavioral
control (Fig. 4). According to TPB, under ideal conditions
and operationalization, the background factors should
affect only the primary antecedents of fertility intentions,
and should not have a direct impact on the intentions
themselves. However, our empirical analysis does not
validate this part of the theory, showing that some of the
background variables also have a direct effect (this is also
in line with Billari et al., 2009). In this respect, for some of
the background variables, the TPB works: this is the case,
for instance, regarding the couple’s education (CEd), as its
influence on fertility intentions is mediated by subjective
norms (SubN). By contrast, a direct influence of the
background variable on fertility intentions has been found
for several demographic factors, such as the number of
previous children (NCh), the woman’s age (AgeW), and the
duration of the couple’s relationship (CDur); for some
socioeconomic factors, such as the woman’s employment
status (WE) and the partners’ housework division (CHD);
as well as for culturally driven factors, such as the degree of
religiosity (Rel). Overall, these results confirm that
stronger equality in terms of division of household labor
favors positive fertility intentions, suggesting that a
perception of fairness of gender arrangements facilitate
Italian women’s plan to have a child in the next period, net
the other factors involved (see also Mills et al., 2008).
Moving to the constraints that may intervene between
the time when people express their fertility intentions and
the moment of the subsequent realization, we considered
the likelihood of a disruption of the couple’s relationship
(Fig. 5, blocks a–d). This event (CDis) is independent of the
level of intentions (FInt) and of all primary antecedents of
fertility intentions (PAtt, NAtt, SubN, PBC), given the
background variables. Among the latters, the woman’s age
(AgeW), the duration of the couple’s relationship (CDur),
the type of couple (CTy), the woman’s employment status
(WE), the woman’s level of satisfaction with the housework division (SHD), and the municipality size (MunS)
have a significant impact on the couple’s disruption risk.
Fig. 5 also offers evidence for the empirical validation of
the TPB. As predicted, having a child (Child) is conditionally
independent of attitudes (PAtt, NAtt), subjective norms
(SubN), and perceived behavioral control (PBC); instead,
these influence the previous step (the formation of
intentions). In other words, the intentions (Fint) act as a
filter between the primary antecedents of fertility plans
and the subsequent behavior. As expected, both the
likelihood of a disruption of the couple’s relationship
(CDis) and the level of intentions (Fint) directly influence
fertility outcomes (Child): the likelihood of having a child
is higher when the fertility intentions are higher and when
the couple remains together in the intervening period (see
Table A3 in the Appendix).
However, the fertility behavior is also directly influenced by some of the background variables, without being
filtered by prior blocks. In particular, the most important
determinants of having a child are the demographics, such
as the woman’s age (AgeW), the number of children (NCh),
and the duration of the couple’s relationship (CDur). This
22
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Fig. 5. Conditional independence graph of background variables (block a), primary antecedents (block b), fertility intentions (block c), actual constraints
(block d), and fertility outcomes (block e). Note: Arrows from block a (background) to block b (primary antecedents) as well as from block a (background) to
block c (constrains) are not shown for simplicity.
has been already stated in previous research (e.g., Noack &
Østby, 2002; Quesnel-Vallée & Morgan, 2003; Rinesi et al.,
2011; Régnier-Loilier & Vignoli, 2011), although without
an examination of the joint distribution of all of the
variables involved in the TPB, as was done in this paper.
Importantly, the couples’ gender role arrangements have a
direct effect for making the step from fertility intentions to
the subsequent behavior. We note that fertility realizations
are naturally higher with stronger intentions, but they also
depend on couple stability. The realizations are also
influenced by demographic variables that are not prefiltered by preceding blocks (i.e., woman’s age, number of
children, and duration of the relationship matter). This
finding confirms that Italian reproductive behavior is
strongly correlated with the demographics, likely being
also drivers of subsequent infecundity and involuntary low
fertility or childlessness.
In sum, the application of graphical models gives
interesting insight to verify if the TPB holds for Italy. The
result is mixed, and can be summarized in Fig. 6. On one
hand, the intermediate variables influence fertility outcomes only by intention (in the above scheme, there is not
a direct link between intermediate variables and child).
This finding supports the TBP. On the other hand, the direct
links between background factors and fertility intentions,
as well as between background factors and fertility
outcomes, are not in line with the TPB.
6. Concluding discussion
We have followed a common paradigm, expecting
individuals to make their procreative choices intentionally.
We relied on a framework built from the TPB (Ajzen, 1991)
and a further adaptation of the TPB to the fertility decisionmaking process (Ajzen, 2010, 2011; Ajzen & Klobas, 2013).
In this paper the TPB is operationalized including fertility
behavior (i.e., the outcomes), while most previous research
applied it for intentions only (Billari et al., 2009;
Dommermuth et al., 2011). Moreover, we moved beyond
existing research by representing all of the possible
dependencies among the variables involved in the TPB,
studied jointly, by means of a chain graph. Our findings
yield new results for the Italian case which are empirically
Fig. 6. Stylized summary of results.
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
robust and theoretically coherent, adding important
insights to the effectiveness of the TPB for fertility research.
Overall, this research represents a contribution to the
study of contemporary fertility decision-making process
from a life course perspective. In this respect, a first key
finding is that attitudes, norms, and perceived behavioral
control are distinct determinants of fertility intentions,
even after adjusting for possible confounding background
factors. And, even more relevant, a second finding is that
none of these dimensions has a direct effect on fertility
behavior: they are all pre-filtered by fertility intentions.
Finally, as a third key finding, we illustrated that the
background factors affect both fertility intentions and
realizations and therefore are not fully mediated by
fertility intentions primary antecedents.
The first two findings are consistent with the theoretical predictions of the TPB and speaks for an effective and
replicable application of TPB in fertility research. We posit
that the distinction between attitudes, norms and behavioral control is a strategy that encourages a simplification
of the complexity of factors leading to fertility behavior.
Our study proposes an approach to the study of fertility
decision making from a life course perspective as a key to
understanding fertility behavior in modern societies.
Importantly, given that similar survey instruments to
the ones used in this paper have been implemented in the
Generations and Gender Survey (GGS) for several other
societies in Europe (Vikat et al., 2007), the approach is also
replicable in a systematic manner for other contexts.
Nevertheless, the third key finding poses a crucial
problem for the operationalization of TPB because of the
role of background factors, complicating the application of
the TPB to fertility research. According to the TPB, under
ideal conditions and operationalization, the background
factors should affect only the primary antecedents of
fertility intentions – namely, attitudes, subjective norms,
and perceived behavioral control – and should not have a
direct impact on the intentions themselves. Contradicting
these assumptions, our findings indicated that, in the
course of the fertility decision-making process, not all
background factors are mediated by the fertility intentions’
primary antecedents. Net of the entire structure of
(in)dependence that characterizes the variables involved
in the TPB, some of the background factors directly
influence fertility intentions, and others even influence
fertility behaviors.
The lack of independence among background factors,
fertility intentions, and outcomes inhibits a complete
validation of TPB for fertility research in the Italian setting.
We located three practical potential reasons for this. First, a
data issue can be at play: it is possible that the battery of
questions provided by the Italian GGS survey do not fully
capture the three conceptually separated dimensions
proposed by TPB. When measurement of the antecedents
of intentions is flawed, related background factors will
compensate and have a direct effect on fertility intentions
and, thus, on fertility outcome (Ajzen, 2010; Ajzen & Klobas,
2013; Dommermuth et al., 2013). Second, our research
design can matter. For pragmatic reasons, due to the smallscale sample, we had to consider all parities jointly
estimating the likelihood of having a child (net of the
23
parity level). Background factors could be parity-specific,
however. We cannot exclude that a future study employing
a sample large enough to stratify the analysis by parity will
bring about different results. Third, a common problem of
all the studies which have tried to empirically test the TPB
(e.g., Dommermuth et al., 2011) regards the use of short
panels, having only two waves. Because the primary
antecedents and the fertility intentions were all collected
at one point in time, the data failed to provide a rigorous test
of TPB: there is no empirical basis for establishing that the
primary antecedents are causally prior to intentions. A
corollary is that some of the background variables used in
the model specification may also be the product of fertility
intentions (e.g., employment, gender role set). Future
research should operationalize the TPB using longitudinal
data with at least three waves in order to insert a time
window not only between the expression of the fertility
intentions and their subsequent outcome, but also between
the surveying of the primary antecedents of intention and
the surveying of the intentions themselves.
A part from data-related reasons, a potential intriguing
and insidious further explanation for the direct effect of
background factors of fertility intentions is theoretical. Our
findings may also venture the existence of further
(psychological) factors which are unobserved, and therefore not built-in the three type of consideration – attitudes,
subjective norms, and perceived behavioral control –
predicted by the TPB. Nevertheless, the link between
background factors and fertility outcomes does not
seriously contrast the TPB, because ‘‘actual enablers and
constraints’’ (placed between the moment of fertility
intentions declaration and their subsequent realizations) is
under-considered in the paper. Due to data limitations, we
only considered the dissolution of the couple as a possible
enabler. The arrays between background factors and
fertility outcomes would probably be statistically irrelevant if we had considered additional intervening constraints before the realization of intentions.
To conclude, it seems relevant to point out that the
analysis could be alternatively carried out using the more
common structural equation models (SEM) instead of
graphical models. Both the classes of models are adequate
for this analysis, even if they are not equivalent (for details,
see Cox & Wermuth, 1993; Smith, Berrington, & Sturgis,
2009). Although less known, graphical models are a wellestablished class of multivariate models, with interesting
probabilistic and computational properties, whose effectiveness and usefulness has still to be explored in social
and demographic studies. The chain graphs utilized in this
paper turns to be not distributionally equivalent to
structural equation model. The advantage in using chain
graphs, even if less known to a general public, is that they
provide a clear interpretation of model parameters in
terms of conditional independence. Such independence
fully corresponds to the missing edges in the corresponding graph. Moreover, none latent variable is introduced in
the model so that no additional distributional assumptions
are required. As a matter of fact, conditional independence
is the key to understand and interpret graphical models,
which seems to us particularly suitable for assessing the
TPB for the fertility decision-making process.
24
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Appendix
See Tables A1–A3.
Table A1
Factor analysis of items of perceived behavioral control, subjective norms, and attitudes regarding the intention to have a/another child within the next
three years.
Items
Factor 1:
Negative
Attitudes
Factor 2:
Positive
attitudes
Factor 3:
Subjective
norms
Factor 4: Perceived
behavioral control
If you were to have a/another child during the next three years, would it be better or worse in relation to. . .
0.58
- The possibility of doing what you want
0.55
- Your employment opportunities
- Your partner’s job opportunities
0.30
- Your financial situation
0.59
0.42
- Your sexual life
- What people think of you
0.41
0.64
- The joy and satisfaction you get from life
- The closeness between you and your partner
0.65
- The closeness between you and your parents
0.55
0.63
- Certainty in your life
If you were to have a child in the next three years, to what extent would the following persons agree with your choice?
0.62
- Most of your friends
- Your mother
0.78
0.71
- Your father
The decision about whether to have children can depend on various situations. How much could your decision about whether to have a child in the next
three years depend on. . .
0.72
- Your economic situation
0.67
- Your job
- Your housing conditions
0.68
- Your health
0.56
0.70
- Your partner’s job
- Help from non-cohab. relatives in caring for the children
0.65
- Help from your partner in caring for the children
0.69
Note: The loadings shown are those that are useful for placing the item in the factor (>.04).
Table A2
Variables considered in the analysis together with descriptive statistics (N = 2871).
Var. name
Block a: background variables
Number of children
NCh
AgeW
Age of woman
CDur
Couple’s duration
CTy
Type of couple
Reg
Region of residence
MunS
Municipality size
CEd
Couple’s levels of education
WE
Woman’s employment situation
Modalities
FREQ (%)
0 (Ref.)
1
2+
<30
30–40 (Ref.)
>40
0–4
5–10 years (Ref.)
>10
Married (Ref.)
Cohabiting
North (Ref.)
Center
South - Islands
Big (Ref.)
Medium
Small
Both low (Ref.)
Both medium
Both high
Her > Him
Him > Her
Public sector (Ref.)
Priv. sect./perm. contr.
Priv. sect./temp. contr.
Not working
12.7
30.3
57.1
9.5
54.2
36.3
15.5
20.9
63.6
96.2
3.8
48.9
17.0
34.1
16.8
39.6
43.6
24.1
29.7
6.1
22.9
17.2
21.0
35.5
4.8
38.8
25
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Table A2 (Continued )
Var. name
ME
Man’s employment situation
CHD
Current housework division
SHD
Woman’s satisfaction with housework division
Rel
Religiosity
Sib
Siblings
Block b: primary antecedents of fertility intentions
PAtt
Positive attitudes
NAtt
Negative attitudes
SubN
Subjective norms
PBC
Perceived behavioral control
Block c: fertility intentions
Fint
Fertility intentions
Block d: constraints 2003–2007
Couple’s disruption
CDis
Block e: fertility outcomes
Birth of a child
Child
Modalities
FREQ (%)
Public sector (Ref.)
Priv. sect./perm. contr.
Priv. sect./temp. contr.
Not working
<95% women (Ref.)
95% women
Yes/moderate (Ref.)
Not at all
At least once per month (Ref.)
Rarely/never
Both partners without siblings
At least one partner with two or more children
Other (Ref.)
19.9
73.8
2.9
3.3
65.6
34.4
4.3
95.7
55.2
44.8
2.4
29.9
67.7
Class
Class
Class
Class
Class
Class
Class
Class
Class
Class
Class
Class
1
2
3
1
2
3
1
2
3
1
2
3
8.8
79.7
11.5
5.4
86.5
8.1
12.0
71.6
16.4
13.9
70.8
15.3
Definitely not
Probably not
Probably yes
Definitely yes
48.7
24.7
16.2
10.5
No (Ref.)
Yes
96.1
3.9
No (Ref.)
Yes
87.2
12.8
Table A3
Detailed model results. Stepwise procedure for model selection with comparison of the reduced model to the full model by means of the Likelihood Ratio
test, to account for the multiple test problem (significance level set at 0.05).
Variables
Categories
Coeff.
Block b: primary antecedents of fertility intentions
Cumulative logit model predicting ‘‘positive attitudes’’ toward childbearing
0 (Ref.)
Number of children
1
2+
Couple’s duration
0–4
5–9 years (Ref.)
10
At least once per month (Ref.)
Religiosity
Rarely/never
Class 1
Negative attitudes
Class 2
Class 3 (Ref.)
Subjective norms
Class 1
Class 2
Class 3 (Ref.)
0
0.205
0.590
0.265
0
0.671
0
0.228
1.471
2.983
0
0.792
2.426
0
Cumulative logit model predicting ‘‘negative attitudes’’ toward childbearing
Number of children
0 (Ref.)
1
2+
Region of residence
North (Ref.)
Center
South
0
0.198
0.408
0
0.091
0.323
St. errors
t-Stat
0.161
0.178
0.162
1.274
3.312
1.636
0.158
4.242
0.099
0.197
0.253
2.306
7.461
11.766
0.153
0.205
5.171
11.852
0.198
0.200
0.998
2.038
0.156
0.129
0.584
2.496
26
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Table A3 (Continued )
Variables
Categories
Current housework division
<95% women (Ref.)
95% women
Class 1
Class 2
Class 3 (Ref.)
Class 1
Class 2
Class 3 (Ref.)
Class 1
Class 2
Class 3 (Ref.)
Positive attitudes
Subjective norms
Pbc
Cumulative logit model predicting ‘‘subjective norms’’ toward childbearing
0 (Ref.)
Number of children
1
2+
<30
Age of woman
30–40 (Ref.)
>40
Couple’s duration
0–4
5–9 years (Ref.)
10
Region of residence
North (Ref.)
Center
South
Both low (Ref.)
Couple’s levels of education
Both medium
Both high
Her > Him
Him > Her
Siblings
Both partners without siblings
At least one partner with large family
Other (Ref.)
Class 1
Positive attitudes
Class 2
Class 3 (Ref.)
Negative attitudes
Class 1
Class 2
Class 3 (Ref.)
Coeff.
0
0.271
1.623
2.906
0
0.713
0.783
0
0.189
0.845
0
0
1.194
2.278
0.189
0
0.391
0.159
0
0.686
0
0.435
0.004
0
0.155
0.609
0.069
0.100
0.631
0.394
0
0.971
2.479
0
0.088
0.697
0
Cumulative logit model predicting the ‘‘perceived behavioral control’’ over childbearing
0 (Ref.)
0
Number of children
1
0.565
0.407
2+
Age of woman
<30
0.351
30–40 (Ref.)
0
0.335
>40
Current housework division
<95% women (Ref.)
0
95% women
0.272
Positive attitudes
Class 1
0.582
0.194
Class 2
Class 3 (Ref.)
0
Negative attitudes
Class 1
0.102
0.978
Class 2
Class 3 (Ref.)
0
Block c: fertility intentions
Cumulative logit model predicting fertility intentions
Number of children
0 (Ref.)
1
2+
<30
Age of woman
30–40 (Ref.)
>40
Couple’s duration
0–4
5–9 years (Ref.)
10
Public sector (Ref.)
Woman’s employment situation
Priv. sect./perm. contr.
Priv. sect./temp. contr.
Not working
0
0.961
2.195
0.278
0
1.125
0.379
0
1.144
0
0.118
0.502
0.218
St. errors
t-Stat
0.122
0.166
0.250
2.230
9.768
11.644
0.165
0.238
4.311
3.289
0.169
0.209
1.115
4.048
0.143
0.161
0.178
8.330
14.193
1.061
0.106
0.163
3.688
0.979
0.172
3.990
0.124
0.102
3.506
0.038
0.124
0.203
0.132
0.141
0.301
0.292
1.249
2.999
0.522
0.710
2.092
1.347
0.155
0.208
6.286
11.940
0.198
0.249
0.445
2.795
0.141
0.137
0.147
4.016
2.958
2.386
0.089
3.753
0.087
0.149
0.193
3.125
3.912
1.004
0.187
0.235
0.546
4.164
0.130
0.143
0.149
7.400
15.384
1.857
0.100
0.139
11.301
2.726
0.147
7.786
0.109
0.202
0.109
1.089
2.487
1.989
27
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Table A3 (Continued )
Variables
Categories
Current housework division
<95% women (Ref.)
95% women
At least once per month (Ref.)
Rarely/never
Class 1
Class 2
Class 3 (Ref.)
Class 1
Class 2
Class 3 (Ref.)
Class 1
Class 2
Class 3 (Ref.)
Class 1
Class 2
Class 3 (Ref.)
Religiosity
Positive attitudes
Negative attitudes
Subjective norms
Pbc
Block d: constraints between 2003 and 2007
Logit model predicting the disruption of couples’ relationships
Age of woman
<30
30–40 (Ref.)
>40
Couple’s duration
0–4
5–9 years (Ref.)
10
Married (Ref.)
Type of couple
Cohabiting
Municipality size
Big (Ref.)
Medium
Small
Woman’s satisfaction
Yes/moderate (Ref.)
Not at all
Regarding housework division
Block e: fertility outcome
Logit model predicting having a child
Number of children
Age of woman
Couple’s duration
Woman’s satisfaction
Regarding housework division
Fertility intentions
Couple’s disruption
0 (Ref.)
1
2+
<30
30–40 (Ref.)
>40
0–4
5–9 years (Ref.)
10
Yes/moderate (Ref.)
Not at all
Definitely not (Ref.)
Probably not
Probably yes
Definitely yes
No (Ref.)
Yes
References
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and
Human Decision Processes, 50(2), 179–211.
Ajzen, I. (2005). Attitudes, personality, and behavior (2nd ed.). Maidenhead,
UK: Open University Press.
Ajzen, I. (2010). Fertility intentions: Theory of planned behavior. Paper presented at BBAW/Leopoldina-Conference on Theoretical Foundations for
the Analysis of Fertility, Lausanne, October 14–16, 2010..
Ajzen, I. (2011). Reflections on Morgan and Bachrach’s critique. Vienna
Yearbook of Population Research, 9, 63–70.
Ajzen, I., & Fishbein, M. (1973). Attitudinal and normative variables as
predictors of specific behaviors. Journal of Personality and Social Psychology, 27, 41–57.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social
behavior. Englewood-Cliffs, NJ: Prentice-Hall.
Ajzen, I., & Klobas, J. (2013). Fertility intentions: An approach based on the
theory of planned behavior. Demographic Research, 29(8), 203–232.
Coeff.
0
0.197
0
0.303
0.893
1.759
0
0.189
1.034
0
0.539
1.306
0
0.361
0.188
0
0.609
0
0.841
1.224
0
0.340
0
3.906
0
1.056
0.035
0
1.853
0
0.466
0.156
0.375
0
1.494
0.885
0
1.553
0
2.207
0
0.647
1.950
3.236
0
1.487
St. errors
t-Stat
0.086
2.298
0.080
0.185
0.219
3.796
4.834
8.015
0.171
0.245
1.103
4.217
0.146
0.180
3.692
7.253
0.120
0.151
3.006
1.246
0.429
1.421
0.460
0.506
1.829
2.416
0.485
0.701
0.369
10.590
0.469
0.408
2.253
0.086
0.517
3.586
0.178
0.226
0.186
2.613
0.691
2.012
0.291
0.179
5.137
4.954
0.212
7.314
0.754
2.926
0.232
0.229
0.250
2.785
8.536
12.935
0.414
3.589
Barber, J. (2011). The Theory of Planned Behavior: Considering dives, proximity and dynamics. Vienna Yearbook of Population Research, 9, 31–35.
Berrington, A. (2004). Perpetual postponers? Women’s, men’s and couples’ fertility
intentions and subsequent fertility behavior. Population Trends, 9–19.
Berrington, A., Hu, Y., Smith, P. W. F., & Sturgis, P. (2008). A graphical chain
model for reciprocal relationships between women’s gender role attitudes and labour force participation. Journal of the Royal Statistical
Society A, 171(1), 89–108.
Berrington, A., & Pattaro, S. (2013). Educational differences in fertility
desires, intentions and behaviour. A life course perspective. Advance
in Life Course Research (on line first).
Billari, F. C., Philipov, D., & Testa, M.-R. (2009). Attitudes, norms and
perceived behavioral control: Explaining fertility intentions in Bulgaria.
, 25(4), 439–465.
Bongaarts, J. (1992). Do reproductive intentions matter? International Family
Planning Perspective, 18(3), 102–108.
Bongaarts, J. (2001). Fertility and reproductive preferences in post-transitional societies. Population and Development Review, 27, 260–281.
28
L. Mencarini et al. / Advances in Life Course Research 23 (2015) 14–28
Borgoni, R., Berrington, A. M., & Smith, P. W. F. (2012). Selecting and fitting graphical
chain models to longitudinal data. Quality and Quantity, 46(3), 715–738.
Bühler, C. (2008). On the structural value of children and its implication on
untended fertility in Bulgaria. Demographic Research, 18, 569–610.
Cavalli, L., & Klobas, J. (2013). How expected life and partner satisfaction
affect women’s fertility outcomes: The role of uncertainty in intentions.
Population Review, 52(2), 70–86.
Cox, D. R. (2008). On an internal method for deriving a summary measure.
Biometrika, 95, 1002–1005.
Cox, D. R., & Wermuth, N. (1993). Linear dependences represented by chain
graphs. Statistical Science, 8, 204–283.
Cox, D. R., & Wermuth, N. (1996). Multivariate dependencies. Models, analysis
and interpretation. London: Chapman and Hall.
Dalla Zuanna, G. (2001). The banquet of Aeolus. An interpretation of Italian
lowest low fertility. Demographic Research, 4(5), 1–35.
De Rose, A., Racioppi, F., & Zanatta, A. L. (2008). Italy: Delayed adaptation of
social institutions to changes in family behavior. Demographic Research,
19, 665–704.
Dommermuth, L., Klobas, J., & Lappegård, T. (2011). Now or later? The Theory
of Planned Behavior and timing of fertility intentions. Advances in Life
Course Research, 16(1), 42–53.
Dommermuth, L., Klobas, J., & Lappegård T. (2013). The theory of planned
behavior and the realization of fertility intentions. Paper presented to the
conference ‘‘Changing families and fertility choices’’, Oslo, 6–7 June 2013..
Edwards, D. (2000). Introduction to graphical modeling (second ed.). New
York: Springer-Verlag.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An
introduction to theory and research. Reading, MA: Addison-Wesley.
Fishbein, M., & Ajzen, I. (2010). Predicting and changing behavior: The reasoned
action approach. New York: Psychology Press.
Frydenberg, M. (1990). The chain graph Markov property. Scandinavian
Journal of Statistics, 17, 333–353.
Goldstein, J., Lutz, W., & Testa, M. R. (2004). The emergence of sub-replacement family size ideals in Europe. Population Research and Policy Review,
22, 479–496.
Gottard, A. (2007). On the inclusion of bivariate marked point processes in
graphical models. Metrika, 66, 269–287.
Heiland, F., Prskawetz, A., & Warren, C. S. (2008). Are individuals’ desired
family size stable? Evidence from West German panel data. European
Journal of Population, 24, 129–156.
Jaccard, J., & Davidson, A. R. (1975). A comparison of two models of social
behavior: Results of a survey sample. Sociometry, 48, 497–517.
Jorgensen, S. R., & Adams, R. P. (1988). Predicting Mexican-American family
planning intentions: An application and test of a social psychological
model. Journal of Marriage and the Family, 50, 107–119.
Kapitány, B., & Spéder, Z. (2012). Realization, postponement or abandonment of childbearing intentions in four European countries. Population,
67(4), 599–629.
Klobas, J. (2011). The Theory of Planned Behavior as a model of reasoning
about fertility decisions. Vienna Yearbook of Population Research, 9,
47–54.
Kuhnt, A. K., & Trappe, H. (2013). Easier to say than done: Childbearing
intentions and their realization in a short term perspective (MPIDR WP
2013-018).
Lauritzen, S. L. (1996). Graphical models. Oxford: Oxford Science Publications.
Lauritzen, S., & Wermuth, N. (1989). Graphical models for associations
between variables, some of which are qualitative and some quantitative.
17, 31–57.
Liefbroer, A. C. (2009). Changes in family size intentions across young
adulthood: A life-course perspective. European Journal of Population,
25(4), 363–386.
Liefbroer, A. C. (2011). On the usefulness of the Theory of Planned Behavior
for fertility research. Vienna Yearbook of Population Research, 9, 55–62.
Mencarini, L., & Tanturri, M. L. (2006). High fertility or childlessness: Microlevel determinants of reproductive behavior in Europe. Population, 4,
389–416.
Miller, W. B. (2011). Comparing the TPB and the T-D-I-B framework. Vienna
Yearbook of Population Research, 9, 19–29.
Mills, M., Mencarini, L., Tanturri, M. L., & Begall, K. (2008). Gender equity and
fertility intentions in Italy and the Netherlands. Demographic Research,
18, 1–26.
Mohamed, W. N., Diamond, I., & Smith, P. W. F. (1998). The determinants of
infant mortality in Malaysia: A graphical chain modelling approach.
Journal of the Royal Statistical Society A, 161, 349–366.
Morgan, S. P. (1982). Parity-specific fertility intentions and uncertainty: The
United Stated, 1970 to 1976. Demography, 19(3), 315–334.
Morgan, S. P., & Bachrach, C. A. (2011). Is the Theory of Planned Behavior an
appropriate model for human fertility? Vienna Yearbook of Population
Research, 9, 11–18.
Neyer, G., Lappegard, T., & Vignoli, D. (2013). Gender equality and
fertility: Which equality matters? European Journal of Population,
29, 245–272.
Noack, T., & Østby, L. (1985). Fertility expectations: A short-cut or dead-end
in predicting fertility? Scandinavian Population Studies, 7, 48–59.
Noack, T., & Østby, L. (2002). Free to choose – But unable to stick it?
Norwegian fertility expectations and subsequent behavior in the following 20 years. In E. Klijzing, & M. Corijn (Eds.), Dynamics of fertility and
partnership in Europe – Insights and lessons from comparative research
(Vol. II). New York/Geneva: United Nations.
Philipov, D. (2009). Fertility intentions and outcomes: The role of policies to
close the gap. European Journal of Population, 25, 355–361.
Philipov, D. (2011). Theories on fertility intentions: A demographer’s perspective. Vienna Yearbook of Population Research, 9, 37–45.
Philipov, D., Spéder, Z., & Billari, F. C. (2006). Soon, later, or ever? The impact
of anomie and social capital on fertility intentions in Bulgaria (2002) and
Hungary (2001). Population Studies, 60(3), 289–308.
Quesnel-Vallée, A., & Morgan, S. P. (2003). Missing the target? Correspondence of fertility intentions and behavior in the U.S. Population Research
and Policy Review, 22, 497–525.
Régnier-Loilier, A. (2006). Influence of own siblings ship size on number of
children desired at various times of life. The case of France. Population,
61(3), 165–194.
Régnier-Loilier, A., & Vignoli, D. (2011). Fertility intentions and obstacles to
their realization in France and Italy. Population-E, 66(2), 361–390.
Rinesi, F., Pinnelli, A., Prati, S., Castagnaro, C., & Iaccarino, C. (2011). Avoir un
deuxième enfant en Italie: de l’intention à la réalisation. Population-F,
66(2), 435–450.
Schoen, R., Astone, N. M., Kim, Y. J., Nathanson, C. A., & Fields, J. M. (1999). Do
fertility intentions affect fertility behavior? Journal of Marriage and the
Family, 61(3), 790–799.
Smith, P. W. F., Berrington, A., & Sturgis, P. (2009). A comparison of graphical
models and structural equation models for the analysis of longitudinal
survey data. In P. Lynn (Ed.), Methodology of longitudinal surveys (pp.
381–392). Chichester, UK: Wiley.
Spéder, Z., & Kapitány, B. (2014). Failure to realize fertility intentions: A key
aspect of the post-communist fertility transition. Population Research
and Policy Review (on line first).
Symeonidou, H. (2000). Expected and actual family size in Greece: 1983–
1997. European Journal of Population, 4, 335–352.
Testa, M.-R, & Grilli, L. (2006). The influence of childbearing regional contexts
on ideal family size in Europe. Population-E, 61(1–2), 109–137.
Thomson, E. (1997). Couple childbearing desires, intentions, and births.
Demography, 34(3), 343–354.
Thomson, E., McDonald, E., & Bumpass, L. L. (1990). Fertility desires and
fertility: Hers, his, and theirs. Demography, 27(4), 579–588.
Toulemon, L., & Testa, M.-R. (2005). Fertility intentions and actual fertility: A
complex relationship. Population et Societies, 415.
Vignoli, D., Rinesi, F., & Mussino, E. (2013). A home to plan the first child?
Fertility intentions and housing conditions in Italy. Population, Space, and
Place, 19, 60–71.
Vikat, A., Spéder, Z., Pailhé, A., Pinnelli, A., Solaz, A., Beets, G., et al. (2007).
Generations and Gender Survey (GGS). Towards a better understanding
of relationships and processes in the life course. Demographic Research,
17, 389–400.
Wright, S. (1921). Correlation and causation. Journal of Agricultural Research,
20, 557–585.