Womens postfertile longevity in Germany, 16581974

1
Womens post-fertile longevity in Germany,
1658-1974
Ola Claësson, Jonas Helgertz and Andrej
Kokkonen. ICPSR Summer Program 2006
Lund University, Göteborg University

• Introduction
• Previous research on factors influencing womens
  post-fertile longevity have focused on parity,
  timing of births and socioeconomic status. Common
  conclusions have been that there is a negative
  relationship between longevity and the number of
  births, suggesting a physical strain of
  childbirth that shortens the womans life span
  (Smith et al, 2002). Results presented by Muller
  et al (2002) furthermore imply that postponement
  of fertility promotes longevity, a finding which
  has been confirmed by several others (see
  references for more citations).
• However, reasonably little effort has been made
  to explore the possible non-biological link
  between longevity and childbearing. Such a link
  would for example be that surviving children
  living close to their mother as adults could
  affect her longevity in a positive way. If she
  had her children close, they could assist her in
  different ways, which would imply that her
  children functioned as an old age insurance. As
  far as we know, this possible link between
  mortality and previous childbirths has never been
  put under investigation before. Below we test
  this hypothesis among others (see square at upper
  right).

Hypotheses H1 There exists a positive
relationship between the amount of social support
and old age longevity. H2 There exists an
interaction effect between the amount of social
support and a womans civil status, which makes
the effect of social support extra pronounced
when a woman is widow (i.e. if she doesnt have a
husband that supports her). H3 There
exists a negative relationship between the
numbers of children ever born and old age
longevity. H4 There exists a positive
correlation between giving birth at late ages and
old age longevity. H5 Differences in
effect of childbearing between socioeconomic
groups found by previous research will disappear
when social support is controlled for.
Conclusions
Data and Method The sample consists of 2774
married and parous women in fourteen German
villages. The data was originally compiled by
John Knodel in XXXX. In our dataset, because we
wanted to investigate the effect of previous
childbearing on mortality, exposure starts at
age 50 for each woman and continues until death.
The reason why there are no censored observations
in our dataset, is because of the criteria that
all women should have been born and reported dead
in the villages that are included in our
investigation. The possible bias in the
investigation due to this criteria can of course
be discussed, since it has possible socioeconomic
implications for the women chosen. This will be
done in the conclusion. An additional criteria
for the women was that they must have been
married and at least have had one child. Our
main principle behind creating the covariate of
interest was to use information about children.
There were three different cases which were
treated in two different ways. Information about
children was taken from the column fate in the
dataset. In the easiest case the child was born
and diseased in the same village. If so, we
considered the child giving social support to his
mother from his fifteenth birthday until the
mothers death. In the two other cases, the child
survived the age of fifteen, but after that his
destiny is somewhat unclear. Either he migrated
at an unknown date or he just didnt appear in
the records anymore. In these last two cases we
have assumed that the child migrated some time
before marriage. We then took the mean age of
marriage (26) for all our women and gave each
child a random out migration date between 15 and
26. In between the period of fifteen and
migration the child was considered to give the
mother social support.
Not a single one of our hypotheses was confirmed
by the analysis! The overall conclusion from
this report is that both SES and Civil status had
a significant impact on longevity for women after
age 50. The civil status showed risks in the
expected direction, reducing mortality with about
11 percent for those still married compared to
those widowed. The independent effect found for
the socioeconomic status, implying higher
mortality for the less privileged, has before
only been found within interaction with number of
births. But no evidence of children functioning
as an old age insurance was found. The
negative results could be a consequence of
selection bias. As we only included married women
who stay in their home villages for their whole
lives we probably bias our sample to include
women with higher SES which dont lack social
support in old age. This suspicion is confirmed
when we look at the distribution of SES. It is
also probable that unobserved heterogeneity
caused our results to become distorted. Firstly,
we omit potentially important variables by not
including controls for the villages women lived
in. Secondly, our data sample stretches over a
time span of more than two hundred years. It is
highly likely that our main variables vary in
effect over this time span. We were, however, not
able to alter our variables to account for such
variations in effect (if they exist) despite
several attempts to do so. We highly recommend
researchers to replicate our study with more
detailed and tailored data. Even studies that
confirm our results would be highly interesting.
After all such results challenge one of the most
cherished ideas in economic demographic theories
that people in pre-modern Europe had children in
order to secure their old age security. Certainly
people would have realized that children didnt
pay off as old age insurance if our results
mirror the reality of pre-modern Europe (and then
we dont even the risk of dying while giving
birth). All things considered children as an old
age insurance was perhaps not such a good idea!
Results In table 2 five models have been
estimated to test our outlined hypotheses.
Throughout the different models, only the total
of three covariates show to significant affect
the risk of mortality for all. In the simple
model, the covariate age at last birth (LB)
implies that the risk of death after age 50
increase with one percent for each year later she
experience her last birth, while the civil status
indicates that those still married have a 12
percent lower mortality than those widowed. In
the model SES model 1 , where socioeconomic
status for the women is added to the simple
model, age at last birth no longer shows
significance. However, the socioeconomic status
have a impact on mortality, increasing in mean
ten percent for each step down towards the less
privileged in the villages. The result from the
three remaining models show that the effects from
SES model 1 are quite stable, and that neither
our social support covariate nor interactions
were even close to show significant effects.


Description and Coding of Variables We did
construct five base variables and two interaction
variables based on four of the five base
variables (basic characteristics of each is shown
in table 1). Our first variable, Children Ever
Born (CEB), measures the number of births each
woman experienced during her lifetime. The second
variable, Age at Last Birth (ALB), measures the
age of the woman at last birth in years. Our
third variable, Civil Status (CS), is a dummy
variable representing married women (value 1).
The fourth variable, Socioeconomic Status (SES),
can take on four values (1-4) of which the lowest
(1) represents the upper classes and the highest
(4) represents the lower classes. Our fifth and
unique variable, Social Support (SS), measures
the number of children living in the same village
as their mother. Children who die or migrate do
thus not count.

References Knodel344060807.mdb database for MS
Access Martin Dribe (2004) Long-term effects of
childbearing on mortality Evidence from
pre-industrial Sweden, Population Studies, vol.
54(3) George Alter, Martin Dribe, and Frans van
Poppel (2006) Widowhood, Family Size, and
Post-Reproductive Mortality A Comparative
Analysis of Three Populations in Nineteenth
Century Europe H. G. Muller, J. M. Chiou, J. R.
Carey, and J.L. Chang (2002) Fertility and Life
Span Late Children Enhance Female Longevity,
Journals of Gerontology Series a-Biological
Sciences and Medical Sciences, vol. 57 Gabriele
Doblhammer (2000) Intergenerational correlation
of effective family size in early Quebec
(Canada), Population Studies, vol. 54 K. R.
Smith, G. P. Mineau and L. L. Bean (2002)
Fertility and Post-reproductive Longevity,
Social Biology, vol. 49