Data and method of analysis Trends in first births

3 Theoretical and practical issues

For the analysis of first births, our data set is extensive by any standard, and as register data go it is particularly rich in information about women's own incomes and their unemployment and educational allowances, as well as about their whereabouts. Nevertheless, like any data set, it represents a compromise between what is desirable in theory and what is possible in practice. In particular, the large data size is bought at the expense of the much larger number of variables and the more accurate dating of non-demographic events in good retrospective survey data. Here are some reflections that explain our practices and our choice of variables.

Any analysis of life-history segments makes unusually strong demands on the character and quality of the data. In principle, we follow women month by month and record some aspects of what happens to them in the current month. For some types of events, there is a delay between a decision to engage in a given type of behavior and the outcome that we record as an event that occurs on a given date. Childbearing is a typical example. A first birth is the end-point a process that may have extended over a longish period. Unless the birth was unplanned, the couple has had to arrive at a decision to start a pregnancy, it may have taken the woman some time to conceive, and there has been a period of gestation. With our type of data we cannot get at this process and at what influenced it while it happened, but we must somehow take into account the fact that it preceded the birth. We would have liked to know a woman's economic situation some nine months or so before any month in which we regard her as "under risk" of having a first birth, but we have had to be content with the income earned and public benefits received in the preceding calendar year. To represent the job situation in her immediate surroundings closely before the current month, we have used information about her most recent home municipality, assigned as follows. We have recorded each woman's municipality of residence at the end of each calendar year. If she has had a change of municipality between the beginning and end of a year, we have allocated the move to the middle of the year. This assigns a home municipality to the woman from the middle of any year to the middle of the next. For any month "under risk" of a first birth, we have then used characteristics of a woman's home municipality thus assigned during the preceding twelve-month period between years' midpoints. We make use of municipal data rather than, say, corresponding information on the national level, because we believe that a woman's perceptions and decisions are dominated by her local environment, and the municipality is the best proxy for it that we can obtain. A woman's home municipality may not coincide with the geographical area in which she can hold a job without moving, but then our data on the employment situation are organized in terms of the municipality in which people are domiciled, not according to where they work. Therefore, our contextual employment variable should reflect adequately the general employment situation that our women perceive. We trust that the essentials of the impacts of income and employment determinants will be picked up adequately by our method of analysis.

The fact that we have a valuable data set does not mean that we could not wish for more. If it were possible, we would in fact have liked to include more information about the woman herself and particularly about her family. Since we are analyzing childless women, we would have liked to know for each woman whether she lived with a man at any time during the years we study, and if she did, we would have liked to know his economic situation as well. The decision to start childbearing depends of course not only on a woman's own employment situation but also on whether she has a partner and on his current situation and his prospects. Unfortunately, Sweden does not have a register that contains the family situation for all inhabitants. We do know whether a woman is married at any time, and if she is we can get similar income information for her husband. However, not many childless couples are married in this country. Most of them live in consensual unions; in fact only about one third of the firstborn in any year have married parents. (Some women who have their first babies have no co-residential partners, but that is less of a problem because only an estimated five percent or so of all newborn children have single mothers.) In this situation, it makes no sense to include a woman's marital status and any information about her husband in the analysis, and we have left it out.

We use our income information primarily to characterize the woman's general situation in the labor market and not really to describe her financial status. We will see the operational consequences in more detail below.


Data and method of analysis Trends in first births

logo70.gif (2450 bytes)

Entry into motherhood in Sweden:
the influence of economic factors on the rise and fall in fertility, 1986-1997

Britta Hoem
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871