There have been large fluctuations in the annual number of births in Britain over the past 50 years. At the 1964 peak, baby boom births were 30% above their 1955 level, but during the subsequent 13 years births fell by 35%. After a recovery during 1978-80, there was another decline which was reversed during 1984. These fluctuations will have a profound effect on the age distribution of the population well into the next century. An earlier exercise had suggested that an econometric model which links the birth rate to developments in the labour market can provide insights into the causes of fluctuations in births and can be useful in forecasting births. The results of these age- specific analyses were more satisfactory on both theoretical and statistical grounds for women in their twenties. Among these women, higher women's real wages depressed their birth rate while higher men's wages had the opposite effect. A relatively large generation size also had a depressing effect on the birth rate of women in their 20s, as suggested by Easterlin. The model was also used to forecast out of the sample period: it generally forecast the turning points correctly, although it appears that it will not do so for the latest turnaround in 1984-5.
It appears then that econometric models have some value in understanding and forecasting the birth rate. The poorer results for older women may reflect the changing mix of women of that age by parity (number of births already achieved), since economic theory based on a framework of sequential decision-making suggests that responses to economic variables may differ by birth order.
We therefore use a sequential framework for childbearing decisions in this paper. We assume that a couple decides whether or not to have a(nother) child in the coming year. By influencing a couple's current resources and expectations about future resources, economic developments affect this decision.
The ultimate number, timing and spacing of children are the outcome of a series of these sequential decisions. A fixed lifetime childbearing plan would be a special case of this more general model.
From this perspective, it is natural to model birth rate by birth order. Thus, our dependent variables are conditional birth rates, or hazard rates; that is, births of a given order relative to the population who could have a birth of that order.
This paper reports some interim results of an econometric investigation using this framework. Our models attempt to identify reactions in the childbearing behaviour of both males and females in response to changes in the labour market. More generally, the models can be used to analyse complex dynamic adjustments within the demographic series to any sort of disturbance.
The variables our models seek to explain are the number of births per woman in a given year. The numerator of these fertility rates distinguishes both the order of the birth (first, second, third or fourth) and the age of the mother. The denominators reflect the number of women of a given age at risk to produce a birth of a given order, that is, women with no previous births are at risk to have first births, mothers with just one child would be those who might have a second, and so on. We felt it important to take account of birth order and age since much of the fluctuation in total birth rates is due to changes in the timing of the births in the (increasingly) standard two-child family: changes in the average ultimate family size have been more gradual. We use data on births occurring between 1952 and 1984 in England and Wales to women born since 1920. Our data set incorporates estimates made by OPCS to reflect the order of illegitimate births.
The economic information included as explanatory variables is primarily focused on the labour market: the ratio of men's to women's wages, the general level of real earnings, price inflation, unemployment, and an index of female attachment to the labour force. We do not use the actual levels of female labour force participation as predictors of fertility, because of the possibility of reverse causation. We use an indicator of each cohort's long-run propensity to engage in paid work, net of the effects on participation of the cohort's current responsibility for childrearing.
We adopt two separate approaches in the analysis of this data set. One approach is designed to trace the influence of economic factors over the life cycle and to detect long-run influences on completed family size using a panel of data with separate observations for each age and year. The other approach tracks relatively short-run adjustments of the fertility series and is designed to forecast a few years ahead. We study such short-run behaviour after first dividing the data into time-series for each (mother's) age group. We have, so far, done this only for first, second and third births. Our preliminary results suggest that it is worth attempting short-run fertility forecasting using time- series models, which produced more accurate short-term forecasts than did the panel data approach. For instance, we used the age- group parameter estimates (based on data up to 1983) along with actual economic data for 1983-5 to forecast age- and order- specific fertility rates and births for 1984 and 1985. The forecast errors for total first, second and third births in 1984 were each less than 1%. While the forecast error for total first births rose to 2.8% in 1985, that only represents an error of 7,500 first births, which is fairly small. The forecast error for second births remained below 1% while that for third births rose to 4.3%. We are less confident about our ability to model long-run influences on family size or to forecast far into the future, although some interesting preliminary results have emerged.
The results confirm some of our previous work on the economic influences on childbearing, in that both approaches find evidence that the opportunity cost of childbearing inhibits fertility at certain stages of the life cycle. Higher wages for women relative to men discourage early entry into childbearing. This subsequently affects the number of higher-order births, if only through reducing the numbers at risk to produce them. Higher attachment to the labour force appears to inhibit the addition of a third child to families of mothers over 35. Mothers with high labour force attachment tend to bunch their first and second births between ages 25 and 29.
Factors which help to speed early family formation in these models (apart from a change in relative wages in favour of men) include the arrival of a relatively small cohort of mothers and low inflation, according to the age-group approach. Rising real earnings which hold constant the male/female wage ratio, unemployment, inflation and generation size all have a variety of generally negative effects. The estimated effects of unemployment are relatively small and ambiguous in direction.
The complicated dynamic structure of our model makes it difficult to trace through the effect of particular economic developments, such as the rise in women's relative pay during the 1970s associated with the Equal Pay Act. In order to do this, we constructed a dynamic simulation model based on the (preferred) parameter estimates from the age-group approach, and simulated this model over the period 1971-85, when there were large fluctuations in births. Given initial values for the birth rates and populations at risk in 1970, the simulation is fully dynamic after 1970, predicting lagged dependent variables and populations at risk. When the economic variables take on their actual values, the dynamic simulation tracks first and second births fairly well.
We use the model to estimate the impact of the changes in women's relative pay during the 1970s on first and second births between 1971 and 1985. Women's real hourly earnings rose from 63% of men's in 1970 to 74% in 1977, after which they fell slightly and were then relatively constant. We simulate what would have happened to births if women's hourly pay had grown at the same rate as men's after 1970, thereby keeping their relative pay constant. The results indicate that in the absence of the changes in women's relative pay, first births would have fluctuated much less between 1973 and 1979 and would have been higher. There also would have been more second births after 1974, in large part because there would have been more first births. There would have been about 130,000 more births between 1973 and 1983 if the ratio of women's wages to men's would have remained constant at its 1970 value. Thus it would appear that changes in women's relative wages substantially reduced the number and altered the timing of births during the 1970s. The simulation does not suggest that the fall in fertility in the 1970s was entirely attributable to the changes in relative wages, but that this factor intensified it.
The results presented here constitute only an interim stage in our study of economic influences on fertility. The set of economic influences examined has subsequently been extended to allow for the effects of taxes and transfer payments and for developments in the housing market. These are fully reported in another paper which also adopts (and develops) the time-series approach. The results of the earlier paper strengthen the findings of the present paper and dispel some of its ambiguities.
The following story emerges from our series of studies of female employment and fertility in postwar Britain. An upward trend in the labour force attachment of successive generations accounts for rising female employment (primarily of middle-aged married women) during the 1950s and 1960s, when fertility was also high and rising. The effects of this trend weakened in the 1970s but the numbers of economically active women were by that time swollen by the fall in the numbers of children and by the fall in the numbers of women entering motherhood in their twenties. This growth in women's propensity to take paid work probably has both economic and social origins, but our analyses of fertility suggest that it is not an important explanatory factor in the baby bust. The ratio of women's to men's pay emerges as the economic factor which contributed to falling birth rates during the 1970s, but in contrast it does not seem to have played much of a direct role in explaining rising aggregate labour supply. The once-and-for-all improvement in women's relative pay associated with the enactment of equal pay legislation seems to have affected the female labour force indirectly by encouraging the deferment of childbearing in the mid-1970s.
We conclude that these preliminary results suggest the desirability of looking beyond the female labour market for an adequate characterization of economic influences on fertility.