We structure our results by setting up models in which we combine both sexes and models which we run separately for female and male respondents.
In models where we use both sexes (Table 3) we control for covariates that are available for each partner separately (age, pre-union children) and covariates that are specific to the current union (marital status, calendar time period, duration of union). To obtain a better understanding of the effect of pre-union children on the intensity of the conception of a first common child in a second union, we investigate different representations of the variable `pre-union children'. In our first model, we distinguish only between a woman's and a man's pre-union children (Table 3, Model 1). In a second step (Table 3, Model 2) we include information on the residence of pre-union children of each partner at the time of the formation of the second union, i.e. whether the child lived in the household at the time of the formation of the second union or not. Our last alternative combines all pre-union children, regardless of whether they belong to the male or female partner, and only distinguishes between pre-union children living in the current household or not living in the household at the time of the formation of the second union (Table 3, Model 3).
The effect coefficients on pre-union children in Table 3 suggest that the intensity of the conception of a first common child in a second union depends essentially on pre-union children living in the household (Model 3) and on whose children one considers (Model 2).
Having more than one pre-union child living in the household at the time of the formation of the second union significantly decreases the intensity of the conception of a first common child as compared to the baseline level of no child in the common household. This result holds irrespective of whether we refer to the total number of pre-union children living in the household (Model 3) or to a gender-specific variable as in Model 2 [Note 15]. This correspondence no longer holds if there is a pre-union child present. The presence of one pre-union child living in the household at the time of the formation of the second union does not effect the rate at which the conception of a first common child occurs (as compared to having no pre-union child living in the household) if it is the woman's child, but it significantly increases the intensity of the conception of a first common child if it is the man's child (Model 2). Later on we show that it is a special combination of a woman's and a man's pre-union children living in the household that determines this high intensity.
As these results suggest, we can represent the variable pre-union children for children not living in the household as a total number of children for the couple but we should use a more detailed definition for pre-union children living in the household (Model 4). For children living in the household at the time of the formation of the second union it is important to distinguish whose children they are.
The effect coefficients on the other covariates included in our regressions are robust across the four models (Table 3). While the woman's age at the start of the second union is significant - the intensity of the conception of a first common child decreases with increasing age - the man's age at the start of the second union does not significantly influence the intensity of a conception. This result supports the general findings that the woman's biological age is an important factor as regards fertility in second unions but that the man's age is not. The intensity of the conception of a first child is twice as high for married couples as it is for couples who are cohabiting. The intensity is higher during the pre-cohabitation time than it is in the cohabitational period, but not significantly. The calendar period of observation turns out not to be significant, but there seems to be some increase in the conception intensity over calendar time.
For the duration variable we observe a non-monotonic shape. The highest intensity of the conception of a first common child is observed from the 10th to the 15th month of observation. For 71.3 per cent of our respondents this period corresponds to the first half year after moving in together. For the period from the 16th to the 21st month, which corresponds to the second half of the first year after union formation for the majority of respondents, the intensity of the conception of the first child does not differ significantly from the baseline level of conceiving a child in the period from the 10th to 15th month. But from the 22nd month (the second year after union formation) onward, we observe a pronounced decline in the intensity of the conception of a first common child. From the fifth year onward, the intensity of the conception of a first common child is only one tenth of the baseline level.
To check the proportionality assumption of the proportional hazard model for model 4 we have tested for interactions between the duration variable and each covariate included. Several interactions turned out to be significant: one with the woman's pre-union children living in the household and the other with the man's pre-union children living in the household. We present the results by referring to a three-way interaction between the man's children, the woman's children and the duration of the second union ([Note 16], Figure 1). Conception during the first nine months, which for the majority of the respondents is the time before moving in together, is highest for couples where the woman has no children and the man will bring one or two pre-union child(ren) into the new household (f0-m1+). These findings might reflect the desire of a childless woman to have a biological child with her partner as soon as possible if he brings a child into the household. Moreover, they show that the characteristics of both partners should be taken into consideration to examine the influence of each partner on the timing of the conception of the first child (see also [Corijn et al. 1996]). Similarly, [Thomson and Hoem 1998] found that couples had a higher risk of birth when the man had a child before the current union than when the man had no children. As a consequence of the high quality of the Austrian data we were able to disentangle the effect of the residence of children and locate that the effect stems from the man's pre-union children living in the same household.
Further interactions that proved to be statistically but not substantively significant are relegated to Appendix A.
To control for characteristics of the preceding union (which are only available for the respondent) we run regressions for female and male respondents separately (Table 4 and Table 5). In fact, running gender-specific models might reveal another important aspect of our set-up that has been neglected so far. Recalling that we can only determine the order of the union for the respondent, the models in Table 3 neglect the fact that women whose second union we consider may be different from women that are partners of men whose second union we consider. It might be a reasonable assumption that men who enter a second union are more likely to have female partners for whom it might be the first union. On the other hand, women who enter their second union might be more likely to have male partners for whom it is already the second or a higher-order union [Note 17]. Hence, gender-specific models may result in differing effect coefficients on gender-specific variables, although we have not yet found any significant influence of the covariate "gender" on the conception intensities in models where both sexes are combined.
Gender-specific models show no significant influence on first conception intensities of the analysed characteristics of the first union, i.e. the dissolution type of the former union and whether or not the respondent was formerly married. We could not find a reduced risk of a first common child if the male respondent was formerly married, as [Heekerens 1986] did for German data and [Griffith et al. 1985] for U.S. data. Nor did we find a higher risk for previously widowed women compared to women whose first marriage ended in divorce, as was found for American data by [Wineberg 1990] [Note 18].
The effect coefficients on the other covariates are pretty much consistent with the results in Table 3 - with some exceptions that might well be caused by the previously mentioned caveat that the behavior of female respondents might differ from that of female partners of male respondents whose second union we consider. Most striking is the fact that among all pre-union children the presence of the man's children in the household have an effect on the conception of the first shared child only for female respondents and not for male respondents. Further investigation shows that, again, this effect is caused by couples with the following characteristics: the woman - whose second union we are concentrating on - is still childless and the man brings one pre-union child into the common household.
The importance of the age of the youngest child for subsequent fertility decisions has been stressed unambiguously in the literature and is verified in our analysis as well (Table 5). Women who have two (or more) pre-union children and whose youngest child is between five and seven years of age have a significantly lower risk of conceiving a first common child in a second union. If a woman has one or more pre-union children all under the age of five, there is no significant age effect of the youngest child nor any significant difference from childless women. Note that the variable "age of youngest child" is organized according to the number of pre-union children. This is in contrast to the findings in [Griffith et al. 1985], where it is independent of the number of children. Furthermore a test on the proportionality assumption reveals an interaction between the age of the youngest child and the duration variable (cf. Appendix B).
Fertility in second unions in Austria: Findings from the Austrian FFS
Isabella Buber, Alexia Prskawetz
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871