Lowest Low Fertility in an Urban
Context: The Role of Migration
in Turin, Italy
Francesca Michielin*
Istituto di Metodi Quantitativi, Universit Bocconi, Viale Isonzo 25, Milano, Italy
INTRODUCTION
F
ertility levels make a significant contribu-
tion to the age structure of the population,
and we need to consider the potential
consequences of low fertility. As Bongaarts and
Feeney (1998: 285) asserted, `declining popula-
tion size would be salutary from some points of
view, but rapid population aging is likely to pose
profound social and economic problems'. The
issue, therefore, becomes particularly important
in lowest-low fertility settings (Kohler et al.,
2002). In this paper, we focus our attention on the
urban context in a lowest-low fertility country
(here represented by the municipality of Turin,
Italy, the inner city of an important metropolitan
area belonging to the `industrial triangle', the
major industrial area in northwestern Italy;
Bonifazi and Heins, 2000), in which fertility
choices appear to be different to elsewhere.
Nowadays, urban total fertility rates (TFRs) in
cities such as Turin, Milan, Udine and Florence
are lower than for Italy as a whole (see Ongaro,
2002). For example, while the Italian TFR for the
year 2000 was above 1.2, the levels for these four
cities ranged between 1.0 and 1.1.
Numerous studies have dealt with urban fer-
tility, mainly focusing on urbanisation processes
in developing countries (see the early review by
Zarate and de Zarate, 1975), while others com-
pared fertility between metropolitan and non-
metropolitan areas in developed countries (see
Courgeau, 1989). In general, it is well known that
urbanisation and industrialisation have pro-
duced many benefits for families and societies,
and at the same time they have exerted pressures
on the family (see United Nations, 1980). For
example, in an urban context, women are more
POPULATION, SPACE AND PLACE
Popul. Space Place 10, 331­347 (2004)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/psp.337
Copyright  2004 John Wiley & Sons, Ltd.
ABSTRACT
In countries with so-called `lowest-low'
fertility, the lowest fertility levels are seen in
the cities. The main reason for this is the
difference in the cost of living, combined with
income constraints in cities, compared with
rural areas. If we focus our attention on the
centre of an urban area, migration needs to be
taken into account, since migrants may have
particularly low fertility levels. In this paper
we use the Turin Longitudinal Study, which
has data on all people who have ever been
residents of Turin (Italy) during the period
1971­2001. We study the interdependencies
between fertility and out-migration choices
for a selected group from the 1956 birth
cohort. Our findings underline the important
role of economic resources and life-cycle
events which seem to guide both fertility and
migration behaviours. Moreover, while having
a child significantly hampers long-distance
migration, it has less impact on short-distance
moves. Copyright  2004 John Wiley & Sons,
Ltd.
Received 3 June 2003; revised 21 October 2003; accepted 28 April
2004
Keywords: out-migration; simultaneous
modelling; Turin Longitudinal Study; urban
fertility
* Correspondence to: F. Michielin, Istituto di Metodi Quanti-
tativi, Universit Bocconi, Viale Isonzo 25, Milano, Italy.
E-mail: francesca.michielin@uni-bocconi.it
likely to participate in the labour market, and
their role within the family may be different to
elsewhere. The cost of living and other income
constraints make living in an urban area more
expensive than living in a rural area, and this
may influence family size (Stark, 1991).
As cities evolve into metropolises, this problem
becomes potentially more serious. As Martinotti
(1993) underlined, many changes in the
morphology of the major cities of Europe have
occurred during recent decades, following the
pioneer experience of the US (Bogue, 1955; Frey
and Speare, 1992). After a period of rapid urban-
isation (which started in Italy during the 1950s
and ended in the mid-1970s), the growth of the
largest cities began to slow down. The popula-
tion of Turin municipality, for example, which
was approximately 700,000 people in 1951,
reached 1,100,000 at the beginning of the 1960s,
mainly due to the evolution of the Fiat car indus-
try. Since the mid-1960s, the central area began to
be characterised by lower (but still positive) net
migration flows, while the outlying districts or
urban ring grew more rapidly. Only during the
early 1970s did Turin municipality begin to
experience net out-migration, while the suburbs
experienced small net inflows (IRES, 1994). These
new migration patterns have been summarised
as `counter-urbanization' (Berry, 1976) or `de-
urbanisation'. As a result, at the beginning of
2001 the population of Turin city was approxi-
mately 900,000, which corresponds to more than
half of the population of Turin's entire metropol-
itan area (consisting of about 1.6 million), and to
two-fifths of the province (2.2 millions).
Studying urban fertility in Northern Italy is
therefore particularly appealing, since we are
dealing with a context of lowest fertility in a
country of lowest low fertility. The aim of the
present article is to analyse fertility levels in the
urban context of Turin municipality (i.e. in
the centre of a relatively large metropolis) as well
as observing how out-migration choices can be
related to fertility behaviour. The analysis also
controls for the possible endogeneity of the
process of urban fertility on out-migration.
Primarily, if we want to understand fertility
behaviour in the centre of an urban area, we also
need to consider the fact that out-migration can
be motivated by the current household situation,
and also by desired fertility. Both processes need
to be taken into account simultaneously.
In the next section we underline the reciprocal
impact of migration on fertility and vice versa,
and look at possible common factors which influ-
ence both decisions. There follows a description
of the data and the types of models that will be
used, and then a discussion of our main findings.
Finally I present my concluding remarks.
MIGRATION AND FERTILITY AS
(POTENTIALLY) INTERRELATED PROCESSES
The study of the interrelationship between fertil-
ity and the migration process has mainly focused
on two different perspectives. On the one hand,
researchers have been particularly interested in
the impact of migration on fertility, studying the
fertility of in-migrants, while on the other hand,
current parity and anticipated fertility have been
considered among the critical determinants for
migration decision-making.
In-Migration and Fertility
Concerning the behaviour of in-migrants, the
literature has focused on testing certain basic
hypotheses which could shed some light on the
mechanisms that influence fertility before and
after migration. The major hypotheses describing
different situations were adaptation, disruption
and selection. Firstly, adaptation (Goldstein and
Goldstein, 1983; Stephen and Bean, 1992) predicts
a model where migrants gradually assimilate to
the fertility norms and behaviour of the host
society. Secondly, the disruption model assumes
that migration will have a temporary effect on
fertility, depressing it shortly after the move,
because of spousal separation or the settling-in
process (Carlson, 1985). Finally, the selection
model (Hervitz, 1985; Kahn, 1994) stresses that
migrants are selected through socioeconomic
characteristics, which in turn also influence
fertility behaviour: controlling for these charac-
teristics should mean that there are no differences
in fertility between migrants and non-migrants.
The literature has mainly tested these three
hypotheses with respect to the urbanisation
process itself by focusing on urban and rural dif-
ferentials in in-migrant fertility, in both develop-
ing and industrialised countries. In addition,
some studies have focused on multicultural
countries, such as the US and Australia, where
consistent international migration flows make a
332 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
significant contribution to overall fertility levels.
For instance, Ware (1975), Carlson (1985) and
Abbasi-Shavazi and McDonald (2000, 2002)
studied the fertility behaviour of Australian
immigrants; Ford (1990), Stephen and Bean
(1992) and Kahn (1994) focused their attention on
the US; while Hervitz (1985) dealt with Brazil;
and Goldstein (1973) and Zarate and de Zarate
(1975) considered the urbanisation process in
developing countries.
In the context of lowest-low fertility, the fertil-
ity of in-migrants is particularly interesting. This
analysis allows for an understanding of the
impact of in-migrants' fertility on the overall
national level. It considers how different fertility
models are, or are not, applicable when the exter-
nal conditions change due to migration.
Fertility and Out-Migration
In the literature concerning migration, particular
emphasis has been given to life-cycle events as
possible determinants of the decision to move
(Courgeau, 1984). This follows the pioneering
idea expressed by Lee (1966) that migration can
be considered as an instrumental behaviour for
achieving specific goals in some other parallel
career. In this respect, the `household career' acts
as a push and pull factor for the decision to move.
The household career can, therefore, constrain
migration decisions (Mulder, 1993). Long (1972)
demonstrated that married couples without
children are more geographically mobile than
married couples with children, whose mobility is
particularly restricted when the children are of
school age. And, of course, if a family is consid-
ering a move, the net family gain will be evalu-
ated, rather than simply the potential personal
gain of the adult who is considering a migration
opportunity (Mincer, 1978). On the other hand,
the need to adjust housing to changes in the
household composition is an important source
of mobility (Grundy, 1986; Baizan, 2002), and
residential mobility can thus be a possible
response to fertility.
The latter push factor can be important in
central urban areas, where spacious single family
dwelling units are often not available and, when
available, may be expensive. Gentrification
processes may also result in rising costs of home
ownership in certain areas (Zukin, 1987). More-
over, related to the expansion of the urban service
economy, some residential buildings have been
converted to service use, resulting in a further
reduction in dwellings.
Common Factors
Besides the direct effects of fertility on migration
decisions, we also need to take into account some
unobserved factors that could potentially influ-
ence both processes at the same time. Speare
(1974) and Landale and Guest (1985) show, for
instance, that residential preferences may play an
important role. According to Mulder and
Hooimeijer (1999), the importance of the resi-
dential environment increases as the family
grows, since married couples, and especially
those with children, increase their financial
investment in the family. As family commitments
grow, the desire for higher quality dwellings may
also increase and, as a consequence, ownership is
preferred to renting.
Rather than minimal requirements for health
and safety, housing preferences reflect the exis-
tence of some commonly held norms. For most
people, housing should preferably be owned by
the occupants, be of an independent structure,
and have sufficient outdoor and indoor space,
given the age and sex composition of the family
(Morris et al., 1976; McAuley and Nutty, 1982).
Cultural and family sequential norms require,
moreover, that the family is residentially stable
before children are born (Baizan, 2002).
All of these elements discourage family
formation in the central city, which remains the
preferred location for young singles and couples
with no children, but loses its attraction during
family formation, childbearing and child-rearing
life stages. Many researchers in the US believe
that housing market conditions, high levels of
crime and segregation all contribute to outflows
from the central city of important metropolitan
areas. This is especially so for families rather
than for single persons (Frey and Kobrin, 1982;
South and Crowder, 1997). Despite the fact that
American cities are different to European cities,
this provides additional support to the argument
that migration can become a strategy for those
who desire or intend to have children. At the
same time `a reluctance or inability to move to
larger accommodations may, in some circum-
stances, depress fertility, and the availability of
housing may affect any relationship between
Fertility and Migration in Turin 333
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
mobility and fertility' (Grundy, 1986: 404). It is
well known that `changes in residence represent
endogenous determinants in an interdependent
system of demographic relevant processes'
(Flöthmann, 1993: 54), but we focus here only
on the interrelationship between migration and
fertility.
In other words, some decisions to migrate are
part of the strategy that might lead to a family
having children. This means that moves are not
only influenced by current household situations,
but also by individual desires for children and by
the importance attached to cultural and social
norms. We might also draw on the ideas of famil-
ism (Sabagh et al., 1969), whereby we expect more
family-oriented households may be more active
in their search for locations that are suitable for
child welfare and family living.
DATA AND METHODS
The Turin Longitudinal Study (TLS) is a longitu-
dinal database (Creeser, 2001) which consists of
register data linked to 1971, 1981 and 1991 census
data. It has been used extensively for epidemio-
logical studies (Costa and Demaria, 1988; Costa
et al., 1994; Faggiano et al., 1994; Kunst et al., 1998)
and, in a limited way, for socio-demographic
studies (Billari et al., 1999).
The data-set covers the entire period starting
from 1971 (the year in which the register was
computerised) to the end of 2000. Therefore, we
have information on all people who have ever
resided in the Turin municipality since 1971.
Being an exhaustive source, the TLS allows us
to consider fertility and migration behaviour for
the entire population we are interested in,
without sample selection problems.
For the present study, we focus our attention
on women born in 1955 and 1956, for whom we
can follow their entire reproductive period (at the
beginning of the 1970s they were 15 years old,
and 44­45 in 2000). Since our main goal is to
study urban fertility and to concentrate on the
impact of forming a family and having children
on migration behaviour, we selected only those
women who were resident in the Turin munici-
pality on their fifteenth birthday, amounting to
11,143 women. In this way we can be assured that
at the (theoretical) beginning of the reproductive
period, all the women were in Turin, and their
fertility choices will have been influenced by that
urban context (as are other behaviours such as
searching for a partner or finding a job). Some of
these women will have moved into Turin during
their childhood, while others will have lived
there all their lives.
For these women we focus attention on the
period following marriage, explicitly selecting
those 7623 women (70% of women of the
1956­1957 birth cohort) who lived in Turin at
least until marriage and following them from
that time. Since fertility in Italy is almost com-
pletely marital, this allows us to capture their
entire fertility history (Castiglioni and Dalla
Zuanna, 1994; Billari et al., 2002). We are also
interested in the first trigger for migration which
for the majority of women will coincide with
marriage (Billari, 2001).
For all women we have information on both
migration and fertility history until censoring
occurs (which corresponds to death, out-
migration or the end of the year 2000). Their
fertility history can be reconstructed by linking
each person to his/her parents, and using both
register and census information. As shown in Fig.
1, out-migration leads automatically to the cen-
soring of the observation and, since we hypoth-
esise that the two processes are linked, the
censoring event may be correlated with the phe-
nomenon under study. One solution for dealing
with this problem, and the existence of other het-
erogeneous factors in the analyses, is to use a
structural equation for event history models. This
allows us to consider more equations simultane-
ously, including in each some unobserved com-
ponent that in principle can also be correlated
with the phenomenon under study (see, for
example, Lillard, 1993; Lillard and Waite, 1993;
Lillard and Panis, 2000).
Since we study two processes (fertility and out-
migration) we will use two simultaneous equa-
tions; the first equation will describe fertility, the
second out-migration. Fertility can be considered
as a process with repeated events, and the `base-
line hazard' (which describes the hazard depend-
ing on the duration of the exposure) can also be
calculated for each birth. In general, we model
the logarithm of the hazard rate as follows:
(1)
where y(t) is a linear spline1
that captures the
impact of the baseline duration on the intensity;
lnh t y t z u t a x b w tkk k j jj i ii
( ) = ( ) + +( ) + + ( ) +Â Â Â e
334 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
each zk(uk + t) denotes the spline representation
of the effect of a time-varying variable that is a
continuous function of t with origin uk. (As an
example, we may consider the age of the woman.
If at the beginning of the exposure period the
woman has age uk, after a period of time t we
know she is uk + t old.) Moreover, we also
consider other time-constant covariates (xj) and
time-varying covariates (wi(t)) whose effect will
be to shift the baseline hazard proportionally.
The final component (e) denotes an un-
observed component which is constant over
time and specific for each unit. If we assign ep
and
eq
to the two components for the processes p and
q, we can write their joint distribution since
it is assumed to follow a bivariate normal
distribution:
(2)
Both the variances (s2
p, s2
q) and the correlation
between heterogeneity terms (rpq, which follows
from the covariance) are estimated in the model.
This particular way of describing each process
allows the inclusion of unobserved factors that
might influence fertility and migration in the
different models. Concerning migration, the idea
of the existence of heterogeneity between in-
dividuals is not new. In 1955, for example,
Blumen, Kogan and McCarthy developed the
mover­stayer model (Blumen et al., 1955), in
e
e
s s
s s
p
q
p pq
pq q
N

Ë
Á
^

~

Ë
^


Ë
Á
^

~

Ë
Á
^

~~ ,
0
0
2
2
which the population was divided into two
groups (a decision that allowed the problem to
be kept mathematically tractable), representing
those who remain permanently in their state of
origin, and those who move during their life.
Later on, extensions of the model considered dif-
ferent (even continuous) heterogeneity distribu-
tions (e.g. Spilerman, 1972; Davies et al., 1982).
Many scholars pointed out the existence of
unobserved heterogeneity, especially in the
fertility process (Gini, 1924; Heckman and
Walker, 1992). In general, it is supposed that dif-
ferences among women in unobserved fecundity
result in unobserved heterogeneity. In modern
societies, where fertility is perceived as a real
choice, we can additionally think about people's
attitudes toward the family, that is, the amount
to which people are actually family-oriented.
We return here again to the idea of familism,
associated now with fertility behaviour (for a dis-
cussion on perverse effects of familistic norms in
the Italian case, see Dalla Zuanna, 2001). In the
present article, however, we are dealing with
urban fertility, and therefore we have to consider
the meaning of `unobserved' in this peculiar
context (heterogeneity does not directly describe
a lower or higher propensity for having children
in general, but of having children in the Turin
municipality). Allowing for the presence of
potentially correlated unobserved heterogeneity,
we can check whether endogeneity also acts
via these components.
Fertility and Migration in Turin 335
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Married,
parity=1,
in Turin
Married,
parity=2,
in Turin
Married,
parity=1,
out of Turin
Fertility
process
Migration
process
Married,
parity=0,
in Turin
Married,
parity=0,
out of Turin
Married,
parity=2,
out of Turin
Married,
parity= ,
in Turin
Married,
parity= +1,
in Turin
Married,
parity= ,
out of Turin
Married,
parity= +1,
out of Turin
n
n
n
n
Figure 1. The processes under study.
In our model, allowing correlation between
unobserved heterogeneity terms, we state that
the two choices (of leaving the city and of having
an additional child) are not independent. In
particular, a positive correlation between
unobserved heterogeneity terms would mean
that, net of the observed characteristics, women
who have a propensity to bear more children also
have a higher propensity to migrate. On the other
hand, those who have less interest in moving will
also have a lower propensity to have children. If
we can interpret heterogeneity in the fertility
process as a measure of the amount to which
people are actually family-oriented, we would
say that more family-oriented people would will-
ingly leave the city than less family-oriented
people, according to Mulder's interpretation
(1993). In this case, the positive effect of fertility
on migration acts also through the unobserved
components, in such a way that whoever desires
more children is also more prone to leave the city.
Similarly, a negative correlation would mean
that if we consider two women with the same
observed characteristics, the one with a higher
propensity to have children would also prefer to
stay longer in Turin. In contrast, people who
would prefer to migrate would also be less prone
to bear children. This is related to how people
perceive the quality of life in the Turin munici-
pality. We might expect that those who like the
city more will see it as a good place for bearing
children, while those who dislike living in
the city may choose not have children, thereby
confirming Grundy's (1986) theory.
Since the estimated correlation coefficient
represents only the net impact of these two
contrasting forces, a null correlation would mean
that these two factors cancel each other out.
The Equation Describing Fertility
Births are repeatable events, but each birth occurs
within a complex decision-making strategy. The
decision-making model used here hypothesises
that women act rationally to realise a plan of
desired family size (Becker, 1981). Since different
strategies can be compatible with the same
number of children, women can choose to act in
different ways (Yamaguchi and Ferguson, 1995;
Rosina, 2001). Therefore, when describing fertil-
ity, information concerning the past needs to be
considered in order to predict future behaviour.
The basic event of interest is a new birth, and
therefore we analyse the hazard of having an
additional child. In fact, we keep the hazard
relative to first parity distinct from transition to
higher parities, since the former event represents
entry into motherhood while, for the others, the
fertility process has begun (unlike Yamaguchi
and Ferguson, 1995, we also consider transition
to first birth). The first baseline is associated with
the length of marriage, and it is possible that a
number of marriages occurred because of a
pregnancy.
Apart from the first child, the length of previous
interval gives additional information for under-
standing subsequent fertility. Murphy (1992:
148), for example, included various possible
meanings of the interval between births:
`physiological difficulties in conceiving, conti-
nuity in terms of contraceptive usage, possible
episodes of spousal separation, low coital
frequency, stable attitudes to appropriate
birth-interval length, and constraining and
socialization factors due to differing educa-
tional and employment histories.'
In this case we also distinguish the first inter-
val between marriage and first birth from the
subsequent intervals which are between births.
We assume that a short interval between two
births is likely to predict a shorter spacing to the
next one (see Yamaguchi and Ferguson, 1995). If
the interval between marriage and first birth is
very short, as in the case of a pregnancy-caused
marriage, it is possible that there was no real
intention to start the reproductive period, and
hence the interval to subsequent children may be
longer.
The age of the woman at the beginning of each
birth interval may also be significant (Marini,
1981). One possible reason is that a woman who
started to be at risk (that means who married or
had a child of parity j) at a young age will have
a long time to conceive an additional child. She
will also have a higher fecundity, and we can
expect a higher probability of progression to a
higher parity. On the other hand, she may decide
to postpone the event, since she has more time to
make a decision.
Generally, it has been found that if a woman is
very young when she has her first child, she will
have relatively short intervals between her births
and a high level of completed fertility (e.g. Hoem
336 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
and Hoem, 1989). The latter phenomenon is
known as the `engine of fertility' (Rodriguez
et al., 1984) and it can be associated with a strong
investment in family formation. This is due to
family building that begins at very early ages, or
to a conscious desire to attain a larger family size
by a certain age (Yamaguchi and Ferguson, 1995).
We model the effect of age at marriage on first
birth, as a linear spline. We distinguish three
groups of women: women who married very
early in their life with respect to other women
(i.e. before age 23), women who marry at
`normal' ages (between age 23 and 26), and
finally women who marry very late (i.e. after age
26). We expect that if a woman has married very
early in her life, the earlier she married the higher
her probability of conceiving a first child. On the
other hand, if a woman married very late, we
expect that the older the woman the lower the
probability of becoming a mother.
For subsequent births, we expect that the older
the woman, the lower her probability of conceiv-
ing a child, and that this effect differs according
to parity. The woman's age at the beginning of
the exposure to bearing the j-th child (i.e. 9
months after the birth of the child of order j - 1)
will therefore be considered as a linear spline,
specific for each order of birth.
We also consider the possibility that the last
pregnancy led to twins, because in this case
women may have a strong wish to wait for a long
time before a new pregnancy (Standberg and
Hoem, 2002). Rosenzweig and Wolpin (1980)
demonstrated, for example, that having had
twins at parity one results in delayed subsequent
fertility, although it has a negligible impact on
completed family size.
All the variables included up to this point refer
to the history of the fertility process, and their
effects are thought to be analogous to other
contexts. The following characteristics are par-
ticularly important in this specific context. We
consider if the woman was an in-migrant, and
we also control for her educational level. Both
variables are interacted with parity.
We expect higher fertility rates for in-migrants
(since most in-migrant women come from
regions in which fertility is higher than in Turin).
This effect may be constant with respect to parity,
as in-migrants maintain their fertility preferences
in the new society, or may these vanish gradually
over time as in-migrants adapt to the host society.
Concerning the educational level, the result is,
in principle, difficult to predict for two reasons.
Firstly, the relationship between educational
level and fertility is itself quite ambiguous, since
it is the result of the balance between the costs of
rearing children and the possibility of doing so
(Becker, 1981). Women with a high educational
level have a higher earnings potential in the
labour market, which in turn increases the rela-
tive cost of children and therefore reduces the
demand for children. Women may spend more
time in education, and this delays their entry into
marriage (Blossfeld and Huinink, 1991), although
it is not clear whether entry into motherhood is
affected directly.
On the other hand, high levels of education are
usually associated with high incomes, which
defines the economic context for fertility. While in
the past, the opportunity costs of childbearing for
women were assumed to more than compensate
for the income effect (and the opposite for men),
in recent years a positive effect of a mother's edu-
cation on fertility has been found, at least for high
parity. For the birth of the first child, Marini (1984)
and Liefbroer and Corijn (1999) demonstrated
that both educational attainment and labour force
participation have a negative impact on women,
which is stronger for entry into motherhood than
for entering a union. Other studies (see, for
example, Hoem and Hoem, 1989; Kravdal, 1992)
pointed out that for the second and the third
parity, controlling for other covariates, women
who have higher education also have higher
relative fertility. The latter effect seems, however,
to disappear when the existence of unobserved
components is taken into account (Kravdal, 2001).
Beyond these general considerations, we also
need to take into account the fact that in the
urban context the availability of economic
resources is more important than elsewhere, and
therefore we can expect that a high level of
education raises the probability of having an
additional child for high parities.
The Equation Describing Migration
Above, we have suggested that current parity
and desired fertility may influence migration
decision-making. However, demographic vari-
ables will also influence migration. In particular,
we focus on how the migration choices of
married couples are related to their fertility.
Fertility and Migration in Turin 337
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
The main variables will therefore deal with the
household situation. The baseline risk will
measure the risk of migrating as a function of mar-
riage duration. The shorter the marriage, the higher
the probability of migration, since the new house-
hold situation is likely to have altered residential
preferences and needs (Mulder, 1993), and since
marriage very often coincides with leaving the
parental home in Italy (Billari, 2001). Delays in
registering the residential changes can cause
strong dependence with duration of marriage.
Residential adjustment may also be necessary
when the current location is no longer suitable to
family size. Thus, we might expect families with
a larger number of children to be more likely to
move for better and cheaper accommodation
(Mulder and Hooimeijer, 1999). On the other
hand, families may be reluctant to move if
children are of school age (Long, 1972).
We also include a variable which records the
current age of the woman. We would expect a
general resistance to migration as age increases.
And, since all the women in our sample are
married, young women may have relatively few
economic resources, making migration difficult.
Landale and Guest (1985), for example, found
that when controlling for both family life-cycle
stages and residential satisfaction, those over 46
years of age were significantly less likely to move
than younger people.
We also consider additional information, such
as whether the women are in-migrants or not, and
distinguish between provenance of in-migration
(Turin province, North-Central Italy or South
Italy and foreign countries). In general we would
expect that those who have moved previously
would be more likely to make subsequent moves,
at least partly because they have fewer ties in the
Turin municipality, but this effect may vary by
provenance.
Women's educational level is also controlled for.
This variable may be related to the probability of
moving, influencing the extent to which people
wanting to move can fulfil their wish and repre-
senting, on the other hand, the possibility of
staying. Moreover, it may happen that women
with high educational levels are more interested
in staying in the city, since only there can they
find a suitable job. Therefore, the effect of educa-
tional level is not easily predictable. Education is
also associated with job position (seniority of the
post within the organization), which will also
influence migration behaviour (Long, 1974; Da
Vanzo, 1981; Sandefur and Scott, 1981).
RESULTS
Table 1 summarises of the explanatory variables
used in the following models; the results are
shown in Tables 2, 3, 4 and 5. In Table 2 we
present the parameters for the fertility process; in
Table 3 and Table 4 we report the analysis for out-
migration (first considering out-migrations as a
whole, and then distinguishing by destination of
the moves); and in Table 5 we consider fertility
and migration as interdependent processes,
allowing also for correlation between the un-
observed components.
In interpreting the parameters, note that a neg-
ative and significant parameter means that, com-
pared with the reference category, the group
considered shows a lower probability of experi-
encing the event; while a positive and signi-
ficant parameter means a higher probability of
experiencing the event.
Fertility
In Table 2 we present the estimated parameters
of the model which considers fertility. The basic
event of interest is a new birth, and therefore we
analysed the hazard of having an additional
child. Since we included in the analysis variables
referring to previous fertility history, such as the
age at previous birth and the length of the inter-
val between marriage and first birth, the effect of
the number of existing children is not significant.2
Concerning the effect of the length of the inter-
val between marriage and the first birth and
between each subsequent birth, we distinguished
between protogenesic and intergenesic intervals.
As expected, the longer the previous interval
between births, the lower the probability of con-
ceiving a new child. Aside from the protogenesic
one, this is true only for intervals longer than 9
months, reflecting the fact that for premarital
conceptions, the lower the interval, the lower the
probability of conceiving again (which is consis-
tent with our hypothesis concerning unintended
fertility, and with the results for Italy obtained by
Rosina, 2001).
We expected the birth of twins to discourage
women from having more children soon after-
wards. This was confirmed to some degree, as the
338 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
results show that if the last pregnancy led to
twins this has a negative effect on subsequent
fertility, but the effect was not significant.
We also considered the age of the woman at
previous childbirth or at marriage, and this was
considered separately for each order of birth.
Similarly to previous results (for example, Hoem
and Hoem, 1989) we found that those who
married young had the highest probability of
having a first child, while women who married
quite late (at least later than 75% of the same
cohort) delayed motherhood.
As far as age at previous childbirth was con-
cerned, we showed that, as expected, the older
the woman, the lower the probability of bearing
a second or, especially, a third child (as in
Murphy, 1992).
The behaviour of in-migrant women is
particularly interesting. At the beginning of their
reproductive period their fertility behaviour
differs significantly from autochthonous women,
as they have a higher likelihood of bearing a first
child, and also a second one. As the number of
children increases, they seem to conform more
closely to Turin's population and these results
would support the adaptation hypothesis as the
in-migrants' fertility preferences become more
similar to the host population over time
(Goldstein and Goldstein, 1983; Bean and
Swicegood, 1985; Hervitz, 1985).
The effect of educational level also varied
according to parity. Compared with those with
low educational levels, whatever their parity,
those who were highly educated had a lower
probability of conceiving a first child (i.e. of expe-
riencing a transition from parity 0 to parity 1):
both the parameters associated with parity 0,
medium level and parity 0, high level are sig-
nificant and negative. Then, a U-shaped effect
emerges for transition to parity 2 and parity 3.
Indeed, among those who already have one
child, those who have a medium educational
level have a lower probability of having an addi-
tional child (the parameter associated with parity
1 medium level is significant and negative), while
women who have a high educational level show
the highest probability of having a second child;
among those who already have 2 children,
women with a high educational level have the
highest probability of having a third child.
Reaching parity 4 or higher is not dependent on
educational level (the parameters associated with
parity >2, medium or high level are not signifi-
cant). These results support the hypothesis that
Fertility and Migration in Turin 339
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Table 1. Explanatory variables for the models concerning fertility and out-migration choices.
Variable Kind of variable Reference category Knots
Fertility process
Current parity Discrete Parity = 1
Age at marriage Continuous Age 23, 26
Age at previous birth Continuous None
Length interval between marriage and Continuous 9 months
first birth
Length interval between previous births Continuous None
Immigrant Discrete Not immigrant, whatever parity
Twin as last parity Discrete No twins
Educational level Discrete Low educational level,
whatever parity
Variance of heterogeneity component
Out-migration process
Current parity Discrete Parity = 0
Current age of the woman Continuous Age 17, 30, 40
Having school-aged children Discrete No school-aged children
Educational level Discrete Low educational level
Provenance of immigration Discrete Not immigrant
Variance of heterogeneity component
Note that in each model there is a baseline risk described through a spline function.
in the urban area, in order to proceed to high
parities, it is particularly important to have
high levels of resources.
Finally, we found that in our model the para-
meter associated with the variance of the hetero-
geneity component was significant, meaning that
women are heterogeneous in respect to their
propensity to have an additional child, and this
propensity is not controlled for through other
observed covariates.
Models for Out-Migration
Out-migration from Turin municipality is
strongly conditioned by demographic events.
Out-migration is most common in the very first
months of marriage (see Fig. 2, representing
the baseline risk for the out-migration process),
which may not be a surprise as residential
adjustment is often a consequence of marriage
(Grundy and Fox, 1985; Mulder and Wagner,
1993). After one year the risk of leaving Turin
municipality declines only slightly with the
length of marriage.
Other demographic events are important
determinants of out-migration (see Table 3). We
see that having one child seems to discourage
out-migration significantly (the parameter is sig-
nificant and negative: compared with those with
no children, those with one child are 88% likely
to leave Turin municipality). However, there are
no significant differences between those with no
children and those with more than one child in
the likelihood of moving away. And, if children
are school-aged, this creates ties with the place of
residence which make out-migration less likely.
Another factor which can discourage mobility
is age: after the age of 30, the probability of
340 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Table 2. Effect of the covariates on the log-hazard of
having an additional child.
Estimate s.e.
Current parity (ref.: parity = 1)
Parity = 2 -0.1135 (0.225)
Parity = 3 -0.2003 (0.398)
Age at marriage (regressor spline for 1st birth)
Slope age 15­23 -0.0160*** (0.001)
Slope age 23­26 0.0028* (0.002)
Slope age >26 -0.0043*** (0.001)
Age at previous birth (regressor spline, no knots)
Effect of age at 1st
birth -0.0043*** (0.001)
on 2nd
birth
Effect of age at 2nd
birth -0.0085*** (0.001)
on 3rd
birth
Length of previous interval (regressor spline)
Interval between 0.0621*** (0.015)
marriage and 1st
birth
<9 months
Interval between -0.0073*** (0.002)
marriage and 1st
birth
>9 months
Interval between previous -0.0166*** (0.003)
births Slope
Immigrant or not (ref.: not immigrant, with any
number of children)
Immigrant, no children 0.2963*** (0.036)
Immigrant, one child 0.1972*** (0.047)
Immigrant, two children 0.0989 (0.112)
Immigrant, higher parity -0.1939 (0.328)
Twins as last parity (ref.: no twins)
Twins = yes -0.2857 (0.292)
Educational level (ref.: low, with any number of
children)
Parity 0, medium level -0.3063*** (0.054)
Parity 0, high level -0.3404*** (0.076)
Parity 1, medium level -0.3231*** (0.059)
Parity 1, high level 0.3302*** (0.094)
Parity 2, medium level -0.1154 (0.131)
Parity 2, high level 0.4615** (0.225)
Parity > 2, medium level -0.3794 (0.292)
Parity > 2, high level -0.7596 (0.648)
Variance of the heterogeneity component
Sigma-fertility 0.4747*** (0.058)
Log-likelihood -66100.9
Note: Asymptotic standard errors in parentheses.
* Significant at 10% level; ** significant at 5% level;
*** significant at 1% level.
-7
-6
-5
-4
-3
-2
0 12 24 36 48 60
Time since marriage (in months)
ln(baselinerisk)
Figure 2. Baseline risks for the migration process.
moving reduces with increasing age (Landale
and Guest, 1985). Young women are also less
likely to move, although since we selected only
married women, young women in the sample are
those who married early. These women may have
had less resources to move with, as Grundy and
Fox (1985) found in England and Wales in 1971.
We also found that those with higher levels of
education had lower probabilities of migrating
away from Turin. Such individuals may be more
oriented to urban ways of life, and may be more
likely to find suitable accommodation in urban
areas, while those with low levels of education
may be forced to move outside the city to find rel-
atively cheaper accommodation. In this respect,
we have to take into account that the women we
selected spent at least the entire period between
age 15 and marriage in Turin municipality, and
consequently they may have had strong ties with
the city.
Finally, past residential history is also impor-
tant. In comparison with those who have been
resident in Turin throughout, in-migrants from
North­Central Italy had a greater propensity to
leave Turin municipality, while originating from
the South hampers migration. The latter result
shows that those who come from the South have
greater incentives to stay longer. A possible
explanation is that those in-migrants have lower
incentives to move to the northern countryside,
where they have no family ties, and a return to
their place of origin involves a `longest-long' dis-
tance move. We might also speculate that the
consistent flows that took place in the 1950s and
1960s from the southern part of Italy towards
Turin set up something like a `southern com-
munity', which women leave less often.
Finally, also in out-migration choices women
appear to be heterogeneous (the coefficient asso-
ciated with the variance of the heterogeneity
component is significant), meaning that although
we controlled in the model for many observed
covariates, some heterogeneity in the propensity
to leave the city remains, which needs to be taken
into account.
When we distinguish by destination (as in
Table 4), the effect of some covariates changes
slightly. As an example, the negative impact of
having one child is only significant for long-
distance migration out of the Turin province.
Also, having school-aged children has a stronger
negative impact on these longer distance moves
than on moves within Turin province.
Also the effect of the provenance of immigra-
tion changes according to the destination of the
move. Indeed, while coming from the South
hampers out-migration to every destination,
coming from Turin province significantly facili-
tates returns, while originating in the North or
Central Italy encourages longer-distance moves.
Models for Fertility and Migration
Table 5 provides the parameters from a model
which considers fertility and migration as linked
processes. In this model, we are controlling for the
selectivity of migration on the fertility process.
Unobserved factors, which we found to be sig-
nificant in the models presented in Table 2 and
Table 3, seem to be only slightly negatively cor-
related, but the correlation coefficient is not sig-
nificant. There is, therefore, only partial support
Fertility and Migration in Turin 341
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Table 3. Effects of the covariates on the log-hazard of
leaving Turin municipality.
Estimate s.e.
Current parity (ref.: parity = 0)
Parity = 1 -0.1226** (0.058)
Parity = 2 -0.0545 (0.077)
Parity > 2 0.0961 (0.129)
Current age of the woman (duration spline)
Slope age 15­17 0.0814*** (0.030)
Slope age 17­30 0.0026*** (0.001)
Slope age 30­40 -0.0031*** (0.001)
Slope age >40 -0.0112*** (0.003)
School-aged children (ref.: no school-aged children)
Has school aged children -0.1857*** (0.066)
Educational level (ref.: low)
Medium level -0.4489*** (0.057)
High level -1.7008*** (0.120)
Provenance of immigration (ref.: not immigrant)
Turin province 0.0827 (0.069)
North­Central Italy 0.1322** (0.061)
South Italy -0.1880*** (0.051)
Foreign countries 0.1277 (0.139)
Variance of the heterogeneity component
Sigma out-migration 0.7145*** (0.109)
Log-likelihood -31356.3
Note: Asymptotic standard errors in parentheses.
* Significance at 10% level; ** significant at 5% level;
*** significant at 1% level.
for the hypothesis that woman who have a
greater propensity to have children in Turin will
also have a greater desire to stay in Turin's
municipality and vice versa. A possible interpre-
tation of this result is that unobserved compo-
nents can in part be an expression of how people
perceive the quality of life in Turin municipality
(or in a specific neighbourhood) in each dimen-
sion of life. As Speare (1974) stated, residential
satisfaction can have an independent effect on
mobility, even when the effects of background
variables (such as duration of residence, age of
the head of the household, city or suburban loca-
tion, being an owner or a renter, and so on) are
taken into account. Following this interpretation,
the better you feel in Turin, the more you want
to have children there, and the less you desire
to move. At the same time, the less you want
children in Turin, the more you desire to leave.
If we examine the effect on the other coeffi-
cients, we can appreciate that the only interesting
change with respect to the model where fertility
choice was considered independent of out-
migration choice (Tables 2 and 3) was on the
parity coefficient in the migration equation. The
significantly lower probability of out-migration
that was associated with parity 1 is no longer
significant, and the apparent trend, that the more
children a person has, the higher the probability
of moving, still remains. Therefore, even if we
control for correlation across unobserved factors,
having children does not seem directly to hamper
out-migration.
CONCLUSIONS
Using data from the Turin Longitudinal Study,
we analysed urban fertility in Northern Italy.
342 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Table 4. Effect of the covariates on the log-hazard of leaving Turin munici-
pality, distinguishing by destination of the move.
Destination
Turin province Other destinations
Estimate s.e. Estimate s.e.
Current parity (ref.: parity = 0)
Parity = 1 -0.0892 (0.069) -0.2091* (0.107)
Parity = 2 -0.0567 (0.089) -0.0218 (0.153)
Parity > 2 0.0774 (0.144) 0.1800 (0.274)
Current age of the woman (duration spline)
Slope age 15­17 0.0875** (0.039) 0.0749 (0.046)
Slope age 17­30 0.0041*** (0.001) -0.0011 (0.001)
Slope age 30­40 -0.0046*** (0.001) 0.0013 (0.001)
Slope age >40 -0.0108*** (0.003) -0.0124** (0.005)
School-aged children (ref.: no school-aged children)
Has school-aged children -0.1556** (0.074) -0.3234** (0.146)
Educational level (ref.: low)
Medium level -0.4214*** (0.068) -0.4815*** (0.096)
High level -1.7403*** (0.144) -1.4920*** (0.198)
Provenance of immigration (ref.: not immigrant)
Turin province 0.1766** (0.078) -0.1936 (0.136)
North­Central Italy 0.0332 (0.072) 0.3366*** (0.110)
South Italy -0.1891*** (0.060) -0.1764* (0.090)
Foreign countries 0.2715* (0.160) -0.3137 (0.295)
Variance of the heterogeneity component
Sigma out-migration 0.7930*** (0.133) 0.8693*** (0.315)
Log-likelihood -23902.3 -9501.8
Note: as Table 3.
Since data on urban fertility are censored by out-
migration, we studied both fertility and migra-
tion together, controlling for the existence of
correlated unobserved components that may bias
results. We selected people who were resident in
the city at least since age 15, studying their
behaviour after marriage. This particular selec-
tion considers people who chose to stay in the
city at least until marriage. This means that
changes in the decision to stay are likely to be
related to family life-cycle stages.
In this context, fertility seems to be particularly
conditionedbytheeducationallevelofthewoman,
which determines the resources for facing new
births more than the rising opportunity costs of
children (see Becker, 1981). The same covariate is
also important for out-migration. People who
have a high educational level may be more ori-
ented to urban ways of life, and the availability of
a high level of economic resources (here consid-
ered through the educational level of the woman)
may enable people to find suitable accommoda-
tion in the city. People with a lower educational
level (and therefore lower economic resources
available) may be forced to move outside the city
to find relatively cheaper accommodation.
Fertility and Migration in Turin 343
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Table 5. Effect of the covariates on the log-hazard of having an additional child and of leaving Turin municipality
when we consider simultaneously the two hazards (i.e. the unobserved heterogeneity terms are potentially
correlated).
Fertility Estimate s.e.
Current parity (ref.: parity = 1)
Parity = 2 -0.1153 (0.225)
Parity = 3 -0.2088 (0.398)
Age at marriage
Slope age 15­23 -0.0161*** (0.001)
Slope age 23­26 0.0028* (0.002)
Slope age >26 -0.0043*** (0.001)
Age at previous birth
Age at 1st
birth -0.0043*** (0.001)
Age at 2nd
birth -0.0085*** (0.001)
Length of previous interval
Protogenesic <9m 0.0622*** (0.015)
>9m -0.0073*** (0.002)
Intergenesic Slope -0.0166*** (0.003)
Being immigrated (ref.: not immigrant)
Immigrant, parity = 0 0.2985*** (0.036)
Immigrant, parity = 1 0.1998*** (0.047)
Immigrant, parity = 2 0.1040 (0.112)
Immigrant, parity > 2 -0.1914 (0.328)
Twins as last parity (ref.: no twins)
Twins = yes -0.2774 (0.290)
Educational level (ref.: low)
0 child, medium level -0.3028*** (0.055)
0 child, high level -0.3210*** (0.077)
1 child, medium level -0.3182*** (0.059)
1 child, high level 0.3525*** (0.096)
2 childr., medium level -0.1080 (0.131)
2 childr., high level 0.4893** (0.226)
>2 childr., medium level -0.3665 (0.292)
>2 childr., high level -0.7296 (0.649)
Note: as Table 3.
Migration Estimate s.e.
Current parity (ref.: parity = 0)
Parity = 1 -0.0641 (0.075)
Parity = 2 0.0665 (0.128)
Parity > 2 0.2836 (0.201)
Current age of the woman
Slope age 15­17 0.0819*** (0.029)
Slope age 17­30 0.0026*** (0.001)
Slope age 30­40 -0.0029*** (0.001)
Slope age >40 -0.0106*** (0.003)
School-aged children (ref.: no school-aged children)
Has school-aged children -0.1782*** (0.067)
Educational level (ref.: low)
Medium level -0.4419*** (0.058)
High level -1.7063*** (0.121)
Provenance of immigration (ref.: not an immigrant)
Turin province 0.0757 (0.070)
North­Central Italy 0.1196* (0.064)
South Italy -0.1922*** (0.051)
Foreign country 0.1237 (0.140)
Heterogeneity component: Variances and covariance
Variance for fertility 0.4761*** (0.058)
Variance out-migration 0.7408*** (0.106)
Correlation -0.1868 (0.158)
Log-likelihood -97456.5
Moreover, parity seems to have an effect on the
choice of moving, as those with more than two
children were more likely to move away from the
city than people with just one child. When the
number of children is high, the ties with the city
created by them (Long, 1972) are compensated by
the need to find suitable accommodation
(Grundy, 1986; Baizan, 2002).
Concerning the behaviour of in-migrants, we
proved their fertility model is different to those
of non-migrants: at least for first parities they
have a higher probability of having an additional
child. Then their behaviour seems to converge to
that of the host population. This supports the
adaptation hypothesis (Goldstein and Goldstein,
1983; Bean and Swicegood, 1985) which states
that in-migrants behave differently from the host
society until adaptation to the host urban society
occurs.
Including the unobserved component in both
processes allowed for the estimation of unbiased
coefficients (Lillard, 1993). Women appear to be
heterogeneous with respect to their propensity to
leave the city and to have an additional child.
Controlling then for possible correlation across
these components, we found a slightly negative
(although not significant) correlation. This par-
tially supports the idea that out-migration may
be perceived as a possible solution to fertility
plans which cannot be completely fulfilled in the
city. This is in line with findings suggesting that
people may adjust the timing of events in the
family life course in accordance with the
availability of appropriate housing. Murphy
and Sullivan (1985), for instance, discussed the
connection between home-ownership and family
stages in Britain, as did Mulder and Wagner
(2001) in the Netherlands and West Germany.
Our research has a number of limitations.
Firstly, we can only consider fertility behaviour
in Turin, but fertility history is censored when-
ever out-migration occurs. Then, the possible
links between fertility and out-migration choices
are inferred only indirectly through correlation
between the heterogeneity components. In other
words, we cannot measure directly the effect of
out-migration on fertility, and we cannot under-
stand whether fertility choices change once
people leave Turin municipality.
Secondly, some important information could
not be used, such as information relating to
the working career. Indeed, only the census
information is available, for those who resided in
Turin municipality at the time of a census. This
means that for many women this information is
missing and therefore useless.
Finally, we focused on the migration behaviour
of married people, but it would be very interest-
ing to include unmarried individuals in the
analysis, as this would allow us to examine
whether marriage is delayed until out-migration
occurs. Unfortunately, as in the case of fertility
history, information is censored by out-migration.
ACKNOWLEDGEMENTS
I want to thank the Research Group on the
Demography of Early Adulthood of the Max
Planck Institute for Demographic Research,
Rostock, and IRES Piedmont for supporting this
research. Maria Cristina Migliore promoted the
idea of studying the fertility of Turin and created
the research network to conduct the project.
Giuseppe Costa and Moreno Demaria provided
access to the Turin Longitudinal Study and
helped with preparing the data. I received advice
and suggestions from Francesco Billari and
Gianpiero Dalla Zuanna. Thanks also to Gunnar
Andersson, Pau Baizan and Hill Kulu, who were
patient enough to read my work, and thanks in
particular to Lucia Coppola with whom I shared
the Rostock research experience. Again, thanks to
Yvonne Sandor and Susann Baker for editing my
English. Finally, I am grateful to anonymous ref-
erees for their valued comments, and especially
to Prof. Boyle for his additional help.
NOTES
(1) A linear spline is a continuous piecewise linear
function, i.e. a function which is linear between
fixed knots. The slopes between couples of knots
have to be estimated by the model. Using a linear
spline to describe the time dependence is therefore
particularly flexible since entirely different shapes
can be represented with the same spline.
(2) However, in another model, where only the parity
was considered, the effect of parity was significant:
the higher the number of previous children, the
lower the probability of having an additional child.
REFERENCES
Abbasi-Shavazi MJ, McDonald P. 2000. Fertility and
multiculturalism: immigrant fertility in Australia,
344 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
1977­1991. International Migration Review 34:
215­242.
Abbasi-Shavazi MJ, McDonald P. 2002. A comparison
of fertility patterns of European immigrants in
Australia with those in the country of origin. Genus
58: 53­76.
Baizan P. 2002. Formation des ménages et migrations:
analyse biographique de trois générations espagnoles.
Série Démographie 16. Bruylant-Academia:
Louvain-la-Neuve.
Bean FD, Swicegood G. 1985. Mexican American Fertil-
ity Patterns. University of Texas Press: Austin, TX.
Becker GS. 1981. A Treatise on the Family. Harvard
University Press: Cambridge, MA.
Berry BJL. (ed.) 1976. Urbanization and Counterurban-
ization. Sage Publications: Beverly Hills.
Billari FC. 2001. The analysis of early life courses:
complex descriptions of the transition to adulthood.
Journal of Population Research 18: 119­142.
Billari FC, Dalla Zuanna G, Demaria M, Migliore MC.
1999. Un record linkage per studiare la fecondit
delle donne residenti a Torino fra il 1970 e il 1994. In
Societ Italiana di Statistica, Atti della XXXIX Riunione
Scientifica. Istituto di Statistica e Matematica, Istituto
Universitario Navale: Napoli; 535­542.
Billari FC, Castiglioni M, Castro Martn T, Michielin F,
Ongaro F. 2002. Household and union formation in
a Mediterranean fashion: Italy and Spain. In Dynam-
ics of Fertility and Partnership in Europe: Insights and
Lessons from Comparative Research, vol. 2, Klijzing E,
Corijn M (eds). United Nations: Geneva; 17­41.
Blossfeld HP, Huinink J. 1991. Human capital invest-
ments or norms of role transition? How women's
schooling and career affect the process of family
formation. American Journal of Sociology 97: 143­168.
Blumen I, Kogan M, McCarthy PJ. 1955. The Industrial
Mobility of Labor as a Probability Process. Cornell
Studies of Industrial and Labor Relations, No. 6.
Cornell University Press: Ithaca, NY.
Bogue DJ. 1955. Urbanism in the United States.
American Journal of Sociology 60: 471­486.
Bonifazi C, Heins F. 2000. Long-term trends of internal
migration in Italy. International Journal of Population
Geography 6: 111­131.
Bongaats J, Feeney G. 1998. On the quantum and
tempo fertility. Population and Development Review
24: 271­291.
Carlson ED. 1985. The impact of international migra-
tion upon the timing of marriage and childbearing.
Demography 22: 61­72.
Castiglioni M, Dalla Zuanna G. 1994. Innovation and
tradition: reproductive and marital behavior in Italy
in the 1970s and 1980s. European Journal of Population
10: 107­141.
Costa G, Demaria M. 1988. Un sistema long-
itudinale di sorveglianza della mortalit secondo le
caratteristiche socio-economiche, come rilevate ai
censimenti di popolazione: descrizione e documen-
tazione del sistema. Epidemiologia e Prevenzione 10:
37­47.
Costa G, Cadum E, Demaria M. 1994. La mortalit
e le differenze sociali a Torino. In Schegge di salute,
D'ambrosio R, Segnan N (eds). Torino; 19­36.
Courgeau D. 1984. Relations between life cycle and
migration. Population 39: 483­513.
Courgeau D. 1989. Family formation and urbanization.
Population: an English Selection 44: 123­146.
Creeser R. 2001. Turin Longitudinal Study. Long-
itudinal Study Newsletter, Centre for Longitudinal
Studies. http://www.statistics.gov.uk/services/
downloads/lsnews24.pdf, No. 24: 2­4.
Dalla Zuanna G. 2001. The banquet of Aeolus: a
familistic interpretation of Italy's lowest low fertil-
ity. Demographic Research 4: 133­162.
Da Vanzo J. 1981. Microeconomic approaches to study-
ing migration decision. In Migration Decision Making.
Multidisciplinary Approaches to Microlevel Studies in
Developed and Developing Countries, De Jong GF,
Gardner RW (eds). Pergamon Press: New York;
90­129.
Davies RB, Crouchley R, Pickles AR. 1982. Modelling
the evolution of heterogeneity in residential mobil-
ity. Demography 19: 291­299.
Faggiano F, Zanetti R, Costa G. 1994. Cancer risk and
social inequalities in Italy. Journal of Epidemiology and
Community Health 48: 447­452.
Flöthmann JE. 1993. Migration and the life course.
Bulletin de Méthodologie Sociologique 39: 45­58.
Ford K. 1990. Duration of residence in the United
States and the fertility of U.S. immigrants. Interna-
tional Migration Review 24: 34­68.
Frey WH, Kobrin FE. 1982. Changing families and
changing mobility: their impact on the central city.
Demography 19: 261­277.
Frey WH, Speare A Jr. 1992. The revival of
metropolitan population growth in the United
States: an assessment of findings from the 1990
Census. Population and Development Review 18:
129­146.
Gini C. 1924. Premiers Recherches sur la Fécondabilité
de la Femme. Proceedings of International Mathematic
Congress 2: 889­992.
Goldstein S. 1973. Interrelation between migration
and fertility in Thailand. Demography 10: 225­
241.
Goldstein S, Goldstein A. 1983. Migration and Fertility
in Peninsular Malaysia: An Analysis Using Life History
Data. Report No. N-1860-AID. Rand Corporation:
Santa Monica.
Grundy E. 1986. Migration and fertility behaviour in
England and Wales: a record linkage study. Journal
of Biosocial Science 18: 403­423.
Fertility and Migration in Turin 345
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Grundy EMD, Fox AJ. 1985. Migration during early
married life. European Journal of Population 1:
237­263.
Heckman JJ, Walker JR. 1992. Understanding third
births in Sweden. In Demographic Applications of
Event History Analysis, Trussell J, Hankinson R,
Tilton J (eds). Clarendon Press: Oxford; 157­208.
Hervitz HM. 1985. Selectivity, adaptation, or disrup-
tion? A comparison of alternative hypotheses on the
effects of migration on fertility: the case of Brazil.
International Migration Review 19: 293­317.
Hoem B, Hoem JM. 1989. The impact of women's
employment on second and third births in modern
Sweden. Population Studies 43: 47­67.
IRES. 1994. Le trasformazioni demografiche della
popolazione di Torino e dell'area metropolitana. In
Schegge di salute, D'Ambrosio R, Segnan N (eds).
Torino; 37­54.
Kahn JR. 1994. Immigrant and native fertility during
the 1980s: adaptation and expectations for the
future. International Migration Review 28: 501­519.
Kohler HP, Billari FC, Ortega JA. 2002. The emergence
of lowest-low fertility in Europe during the 1990s.
Population and Development Review 28: 641­680.
Kravdal O. 1992. The emergence of a positive relation
between education and third birth rates in Norway
with supportive evidence from the United States.
Population Studies 46: 459­475.
Kravdal O. 2001. The high fertility of college educated
women in Norway: an artifact of the separate mod-
elling of each parity transition. Demographic Research
5: 187­215.
Kunst AE, Groenhof F, Mackenbach JP, EU Working
Group on Socioeconomic Inequalities in Health.
1998. Occupational class and cause specific mortal-
ity in middle aged men in 11 European countries:
comparison of population based studies. British
Medical Journal 316: 1636­1642.
Landale NS, Guest AM. 1985. Constraints, satisfaction
and residential mobility: Speare's model reconsid-
ered. Demography 22: 199­222.
Lee ES. 1966. Theory of migration. Demography 3:
47­57.
Liefbroer AC, Corijn M. 1999. Who, what, where,
and when? Specifying the impact of educational
attainment and labour force participation on family
formation. European Journal of Population 15: 45­
75.
Lillard LA. 1993. Simultaneous equations for hazards:
marriage duration and fertility timing. Journal of
Econometrics 56: 189­217.
Lillard LA, Panis CWA. 2000. aML User's Guide
and Reference Manual. Econ Ware: Los Angeles,
California.
Lillard LA, Waite LJ. 1993. A joint model of marital
childbearing and marital disruption. Demography
30: 653­681.
Long LH. 1972. The influence of number and ages
of children on residential mobility. Demography
9: 371­382.
Long LH. 1974. Women's labor force participation and
the residential mobility of families. Social Forces 52:
342­348.
Marini MM. 1981. Effects of the timing of marriage and
first birth on fertility. Journal of the Marriage and the
Family 43: 27­46.
Marini MM. 1984. Women's educational attainment
and the timing of entry into parenthood. American
Sociological Review 49: 491­511.
Martinotti G. 1993. Metropoli. La nuova morfologia sociale
della citt. Il Mulino: Bologna.
McAuley WJ, Nutty CL. 1982. Residential preferences
and moving behavior: a family life-cycle analysis.
Journal of Marriage and the Family 44: 301­309.
Mincer J. 1978. Family migration decisions. Journal of
Political Economy 86: 749­773.
Morris EW, Crull SR, Winter M. 1976. Housing norms,
housing satisfaction and the propensity to move.
Journal of Marriage and the Family 38: 309­320.
Mulder C. 1993. Migration Dynamics: A Life Course
Approach. Thesis Publishers: Amsterdam.
Mulder C, Wagner M. 1993. Migration and marriage in
the life course: a method for studying synchronized
events. European Journal of Population 9: 55­76.
Mulder C, Hooimeijer. 1999. Residential relocations in
the life course. In Population Issues: an Interdiscipli-
nary Focus, van Wissen LJG, Dykstra PA (eds).
Kluwer Academic / Plenum Publishers: New York;
159­186.
Mulder C, Wagner M. 2001. The connections between
family formation and first-time ownership in the
context of West Germany and the Netherlands.
European Journal of Population 17: 137­164.
Murphy M. 1992. The progression to the third birth in
Sweden. In Demographic Applications of Event History
Analysis, Trussell J, Hankinson R, Tilton J (eds).
Clarendon Press: Oxford; 141­156.
Murphy MJ, Sullivan O. 1985. Housing tenure and
family formation in contemporary Britain. European
Sociological Review 1: 230­243.
Ongaro F. 2002. La bassa fecondit in Italia tra fattori
esplicativi e implicazioni socio-economiche: conseguenze
per la ricerca. Paper presented at XLI Riunione
Scientifica della Societ Italiana di Statistica, Milano
5­7 Giugno 2002.
Rodriguez G, Hobcraft JN, McDonald JW, Menken J,
Trussell J. 1984. Comparative Analysis of Determinants
of Birth Intervals. World Fertility Survey Compara-
tive Studies No. 30. International Statistical Institute:
Voorburg.
Rosenzweig MR, Wolpin KI. 1980. Life-cycle labor
supply and fertility: causal inferences from house-
hold models. Journal of Political Economy 88: 328­
348.
346 F. Michielin
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)
Rosina A. 2001. Analisi delle storie riproduttive. Metodolo-
gia e applicazioni a dati italiani. CLEUP editrice:
Padova.
Sabagh G, Van Arsdol MD, Butler EW. 1969. Some
determinants of metropolitan residential mobility:
conceptual considerations. Social Forces 48: 88­98.
Sandefur GD, Scott WJ. 1981. A dynamic analysis of
migration: an assessment of the effects of age, family
and career variables. Demography 18: 355­368.
South SJ, Crowder KD. 1997. Residential mobility
between cities and suburbs: race, sub-urbanization,
and back-to-the-city moves. Demography 34: 525­538.
Speare A Jr. 1974. Residential satisfaction as an inter-
vening variable in residential mobility. Demography
11: 173­188.
Spilerman S. 1972. Extensions of the mover­stayer
model. American Journal of Sociology 78: 599­626.
Standberg M, Hoem JM. 2002. Patterns of Twinning for
Swedish Women, 1961­1999. MPIDR Working Paper
WP2002-031, Max Planck Institute for Demographic
Research, Rostock, Germany. http://www.demogr.
mpg.de/
Stark O. 1991. The Migration of Labor. Baxil Blackwell:
Cambridge, MA.
Stephen EH, Bean FD. 1992. Assimilation, disruption
and the fertility of Mexican-origin women in the
United States. International Migration Review 26:
67­88.
United Nations. 1980. Patterns of Urban and Rural
Population Growth. United Nations: New York.
Ware H. 1975. Immigrant fertility: behaviour and
attitudes. International Migration Review 9: 361­378.
Yamaguchi K, Ferguson LR. 1995. The stopping and
spacing of childbirths and their birth-history
predictors: rational-choice theory and event-history
analysis. American Sociological Review 60: 272­
298.
Zarate A, de Zarate A. 1975. On the reconciliation of
research findings of migrant­nonmigrant fertility
differentials in urban areas. International Migration
Review 9: 115­156.
Zukin S. 1987. Gentrification: culture and capital in
the urban core. Annual Review of Sociology 13:
129­147.
Fertility and Migration in Turin 347
Copyright  2004 John Wiley & Sons, Ltd. Popul. Space Place 10, 331­347 (2004)