Trends in first births The effect of labor-market attachment

5 The impact of employment trends

To better understand how the local economic environment influences first births, we have used the trend in the per cent employed among men and women at ages 16-64 in a woman's home municipality, measured in November each year. This is the only municipal-employment indicator available to us, but we also regard it as adequate for our purpose, for it should reflect general economic developments as perceived by local couples. To get comparability across municipalities, we have divided each municipal employment rate by the corresponding rate for 1985 (so the index value for 1985 is 1 in all municipalities). This avoids the problem that in any year, employment rates vary strongly from one municipality to another for structural reasons. In a given year, a municipality with a given low employment rate might have low first-birth rates because times are bad, but again it might have high birth rates because it largely covers a rural district where a lower fraction of the population is recorded as having a job. For example, an employment rate of 75 percent may be observed when times are good in municipalities that generally have low employment rates, and the same percentage may be found when times are bad in municipalities that generally have high employment rates. In 1990 (a year with extremely little unemployment) only five municipalities had employment rates below 75 percent and many other municipalities had rates over 90 percent. During the hard times five years later, only one third of the municipalities had employment rates above 75 percent and not a single municipality had employment above 85 percent. This makes it less satisfactory to rely on the municipal employment rates themselves as determinants of fertility than to use their trends, as we have done.

In a first experiment with our employment-trend indicator, we have included it as another covariate in a hazard regression that also includes the calendar year and the woman's age. This gives the interesting result that the strong drop in first-birth rates after 1990 almost disappears in a plot of period effects [Figure 9]. It seems that our trend indicator for the municipal employment level reflects features that "explain" most of the recent decline in first births. The operating element need not be the direct monetary effect of the local employment trends themselves; it seems equally likely that the trends influence childbearing behavior via the impressions couples get concerning how things may develop for themselves in the near future. A sudden drop in local employment may cause a shock effect where the initiation of childbearing is postponed until times take a better turn. Whatever the ingredients of the explanation are, evidently it is not complete, however, for the fertility increase in the late 1980s remains unaffected by the inclusion of the new variable, and some fertility decline remains in our last couple of years.

The relative "risks" on our trend indicator (Column 1 in [Table 1]) show that first-birth rates are strongly influenced by municipal employment trends and move closely in step with them. As we could expect, young women are much more strongly affected than older women are (see [4], Figure 16; not shown here). Much more easily than older women, young women can postpone entry into motherhood when times are unfavorable to childbearing.

Despite our preference for representing a woman's local economic environment by our index for the employment trend in her home municipality, we have also carried out a parallel analysis using the annual employment rates instead. This did not make much difference, except that this alternative variable eliminated less of the trend in first births in recent years. Other combinations of the variables available to us gave intermediary results.

Our employment-trend indicators are so closely related to the period factor that it is hard to keep both in the same analysis. A numerical iteration process is used to estimate our factor effects, and it is difficult to make this process converge when both of these factors are included in the model at the same time. When we leave out the period factor, nothing much happens to the effects we are most interested in, as is seen for instance in the second column of [Table 1]. Therefore, the calendar year is omitted in all subsequent work that involves local employment trends.


Trends in first births The effect of labor-market attachment

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