Transcript Les relations amoureuses stables non cohabitantes. Qui

From fertility intentions to realization:
Implementing Theory of Planned Behaviours
with graphical models
Letizia Mencarini (Università di Torino & Collegio Carlo Alberto)
Daniele Vignoli (Dipartimento di Statistica – Università di Firenze)
Anna Gottard (Dipartimento di Statistica – Università di Firenze)
Objective
• Study of fertility intentions  central on the
discussion of fertility rates
– realistic assumption that, in almost perfect contraceptive
regimes, having a child can be considered a result of a
reasoned decision.
– persistent low fertility levels, especially in Southern Europe,
seem more and more the result of a “fertility gap”
between desired and realized level of fertility
– People declare to want to have more children than they
have but they are for some reasons unable to implement
their desire.
Objective
• This creates quest for research on 
the passage between the intention of having
a(nother) child and the subsequent realization of such
intention.
– plethora of recent studies on determinants and mechanisms
of fertility intentions formation, much less on the successive
step of the realizations
• Italian context particular relevant for this research
question
Theory of Planned Behaviours (TPB)
Evidence from previous studies
•
On the basis of the literature review outlined so far is clear that the basic
background determinants of realizations are generally the same of intentions –
Thompson et al, 1990; Symeonidou, 2000Thomson, 1997; Menniti, 2005, Bühler, 2008;
Meggiolaro, 2009; Rinesi, 2009.
•
But, there are also (obvious) factors that influence only realization and not
intentions (e.g., the degree of intentions, partners’ agreement/disagreement of
fertility intentions) – Thomson et al, 1990; Thomson, 1997; Regnier-Loilier and Vignoli, 2011.
•
Other factors seem to play a stronger role on the realization than on the
intentions: this is the case of women’s age that likely stands as proxy of
infecundity – Noack and Østby, 2002; Quesnel-Vallée and Morgan, 2003; Testa and Toulemon,
2006.
•
And there are intervening factors that can delay or change fertility intentions
such as couples’ problems (i.e. a separation) or other constraints.
The Italian case: striking evidence
•
From DATA: 2003 Family and Social Subjects Survey & its 2007 follow up
» 5742 individuals (in 2871 married and cohabiting couples)
» Expressed fertility intentions for the next 3 years in 2003
» Realisation of 2003 intentions through 2003-2007 fertility (368
children born)
•
As previous studies have already shown
•
We confirm that
•
 Negative fertility intentions are potent predictor of subsequent fertility behavior
•
 Positive fertility intentions tend to overestimate fertility realizations
(e.g., Westoff/Ryder, 1977; Monnier, 1989; Schoen et
al, 1999; Symeonidou, 2000; Noack/Østby, 2002; Toulemon/Testa, 2005; Meggiolaro, 2007; Rinesi, 2009)
The link between positive/negative
intentions & realisations
DATA:
Implementing TPB with 2003 FSS data
•
Questions on intentions that cover an overseeable time period are “in close temporal
proximity to the prospective behavior” (Ajzen and Fishbein 1973: 49):
» “Do you intend to have a child in the next three years?” : “Definitely not”,
“Probably not”, “Probably yes”, and “Definitely yes”.
•
Ten items are available to characterize attitudes towards having a child.
» ‘‘Let us assume that you will have a child or another child in the next 3
years. Would it be better or worse in relation to…’’
•
Three items are available to characterize subjective norms
» ‘‘If you were to have a child in the next 3 years, how much would the
following persons agree with your choice?’’
•
Ten items to capture the perceived behavioral control.
» ‘‘The decision whether to have children or not can depend on various
situations. How much could your decision whether to have a child or not in
the next 3 years depend on...’’
Items
Factor 1
Negative
attitudes
Factor 2
Positive
attitudes
Factor3
Subjective
norms
Factor 4
Perceived
behavioral
control
Let us assume that you will have a child or
another child in the next 3 years. Would it be
better or worse in relation to…
The possibility of doing what you want
Your employment opportunities
Your partner’s job opportunities
Your financial situation
Your sexual life
What people think of you
The joy and satisfaction you get from life
The closeness between you and your partner
The closeness between you and your parents
Certainty in your life
If you were to have a child in the next 3 years,
how much would the following persons agree
with your
choice?
Most of your friends
Your mother
Your father
The decision whether to have children or not can
depend on various situations. How much could
your
decision whether to have a child or not in the next
3 years depend on...
Your economic situation
Your job
Your housing conditions
Your health
Your partner’s job
Help from non-cohabitant relatives in caring for the
children
Help from partner in caring for the children
0.58
0.55
0.30
0.59
0.42
0.41
0.64
0.65
0.55
0.63
0.62
0.78
0.71
0.72
0.67
0.68
0.56
0.70
0.65
0.69
Dimensions of
TPB with FSS data
METHOD:
Implementing TPB with graphical models
•
Graphical models are a class of multivariate models useful to study,
estimate, describe and visualize the relationships among an entire set of
variables of interest.
•
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 ( i.e., nodes represent
variables and the absence of connection between two nodes represents a
conditional independence).
METHOD:
Implementing TPB with graphical models
•
Chain graph models (Lauritzen and Wermuth, 1989) are particularly useful
whenever variables admit a partial ordering on the basis of subjectmatter considerations.
•
Variables are then partitioned into blocks.
» Variables belonging to a same block can be connected
by undirected edges and are considered to be on an
equal standing.
» Variables belonging to different blocks can be joined
by arrows, representing a directed, “causal-like",
association.
•
This class of graphical models seems the most appropriate in order to
empirically implement the TPB.
Structure of associations among
variables in graphical models
•
Two variables, X and Y are independent given a third variable whenever, controlling for Z,
X does not provide additional information on the distribution of Y and vice versa (i.e.,
graphically this means: absence of connection between two nodes).
EXAMPLE:
The lack of the edge between node 1 and 2,
corresponds to the marginal independence
statement X1  X2, as no previous blocks
are present.
On the other hand, the absence of between node
3 and 4, state that X3  X4 | X1;X2.
The absence of the arrow form node 3 to node 5
implies that X3  X5 | X1;X2;X4;X6, as the
node 5 pointed by the missing arrow is in the
last block.
Ordering of variable blocks
in TPB
•
TPB suggests in itself the temporal sequence for the process leading to
the decision to have a child.
•
The sequence is produced by:
1. “background variables” (block a),
2.“perceived behavioral control”, “subjective norms”,
“positive and negative attitudes” (block b),
3.“fertility intentions” (block c),
4.“actual constraints” (block d), and
5.“fertility outcome” (block e).
RESULTS:
Structure of the associations between background
variables and peoples’ own consideration
RESULTS:
Structure of the associations among background
variables, peoples’ considerations, and fertility intentions.
RESULTS:
Structure of the associations among background
variables, peoples’ considerations, fertility intentions,
actual constraints, and fertility outcome.
Concluding remarks
•
The data used (2003 FSS) and the methodological approach followed
(graphical models) allowed a full implementation of TPB
•
Our results validates much of TPB for the Italian context
•
The most important determinants of the transformation of fertility
intentions in subsequent outcomes results to be the proximate
determinants (mainly the demographic ones)
•
The determinants of intentions are not necessarily the same of those of
realisation. It seems to emerge the role of some constraints:
» The gender role-set is crucial for making the step from fertility
intentions to the subsequent behaviour
Thank for your attention!
Letizia MENCARINI:
[email protected]
Daniele VIGNOLI:
[email protected]
Anna GOTTARD:
[email protected]
(2007-2003)
Realisations of positive fertility intentions
% realisations of positive fertility intentions
BY partners’ agreements
% realisations of positive fertility intentions
BY degree of intentions
Factors affecting realisations
(the role of intentions)
With respect to the baseline person
0.447
Partners' agreement over fertility
intentions (Ref. = Agreement)
Disagreement
0.282***
Degree of fertility intentions
(Ref. = probably yes)
Definitely yes
0.727***
Note: Probit marginal effects
Factors affecting realisations
(demographics and couples’ stability)
With respect to the baseline person
Number of children (Ref. = 0)
0.447
1
0.601***
2+
0.624**
<30
0.517***
>40
0.224***
0-4
0.622***
≥10
0.207***
Type of couple (Ref. = Married)
Cohabiting
0.425
Couples' disruption (Ref. = No)
Yes
0.116***
Age of women (Ref. = 30-40)
Couples' duration (Ref. = 5-9 years)
Note: Probit marginal effects
Factors affecting realisations
(area of residence)
With respect to the baseline person
Region of residence (Ref. = North)
Municipality size (Ref. = Big)
Note: Probit marginal effects
0.447
Centre
0.414
Mezzogiorno
0.303***
Medium
0.296***
Small
0.294***
Factors affecting realisations
(couples’ socio-economic situation)
With respect to the baseline person
0.447
Couples' level of education
(Ref. = Both low)
Both medium
Both high
Her > Him
Him > Her
0.522
0.620**
0.563**
0.502
Women's employment situation
Priv. sect. / perm. contr.
0.552**
(Ref. = Public sector)
Priv. sect. / temp. contr.
Not working
0.527
0.587**
Men's employment situation
Priv. sect. / perm. contr.
0.403
(Ref. = Public sector)
Priv. sect. / temp. contr.
Not working
0.508
0.460
Note: Probit marginal effects
Factors affecting realisations
(couples’ role-set)
With respect to the baseline person
0.447
Current housework division
(Ref. = <90% women )
≥ 90% women
Women's satisfaction on housework
division (Ref. = full / moderate
satisfaction)
Dissatisfaction
Note: Probit marginal effects
0.498
0.135***
Factors affecting realisations
(Economic difficulties and parental proximity)
With respect to the baseline person
0.447
Housing tenure (Ref. = tenancy / rent
free)
Home-ownership
0.399
Perception of economic situation
(Ref. = good / medium)
Bad
0.252**
Parents' residential proximity
(Ref. = mother and mother in law both far)
Note: Probit marginal effects
At leat one at medium
distance
At least one at close
distance
0.615***
0.693**
Robustness check for unobserved selection
•
The null Heckman model (intentionsrealisations) displays
significant unobserved selection
•
The complete Heckman model doesn’t show serious problem of
unobserved selection (the correlation coefficient is “on the border” of
significance)
•
However, some modifications shows up:
» The role of parity disappears
» The negative effect of age increases for those
women aged 40+
» The role of the geographical area levels off
Factors affecting realisations
•
Very important “direct” determinants of realisations
» Age
» Degree of intentions
» Couple’s agreement on intentions
•
The “proximate” determinants of realisations are similar to the
ones of intentions
Among the others:
» The role of the perceived gender role-set
Differences between factors affecting the
intentions & realisations
•
On intentions BUT NOT on realisations
» Religious services attendance
» Siblings
•
On realisations BUT NOT on intentions
–
(neg. eff. of ) Perception of future economic constraints
–
(neg. eff. of ) Residence on small/medium municipalities (proxy of
services?)
–
(pos. eff. of ) Residential proximity with grandparents