Women’s Wages and Fertility Revisited. Evidence from a Country with Good Data

Tom Kornstad, Research Dep., Statistics Norway

According to Beckers’ New Home Economics (NHE) females'' wage rate is one of the most important determinants of fertility as it reflects the opportunity cost of having children. The prediction from theory of a negative effect of female wages on fertility has been tested in a number of studies, but the results are diverging. We contribute with new evidence based on registry data covering all Norwegian women born 1955–74 and a simultaneous hazard model of transitions to first, second and third birth. So far we have found a U-shaped relationship between hourly wages and the log hazard for all the four cohorts we are studying, however varying in strength and across parity. However, as women are located at different points on the U-shaped curve, it hard to summarize the total effect of wages on womens'' fertility. To fully understand this relationship, it would be preferable to use the estimated model for simulations. Then we can quantify the relationship between womens'' ages and their probability of giving birth to child. We intend to study the following situations: 1) Suppose all women get a 10 perecent increase in the wage rate. What are the effects on transitions to first, second and third birth? 2) One might suspect that the effect of wage changes is dependent on the woman''s location in the wage distribution. For instance, the effect might differ for women in the lower part of the wage distribution compared to women in the upper part of the distribution. To study this question we intend to divide the women into three separate groups according to wage, and undertake the simutions separately for each of these groups. As the analysis divides the women into four birth cohorts, we can also study how the effect varies over time.

Beckers’ New Home Economics (NHE) is one of the most influential and formally coherent theory of fertility behaviour (Becker 1960, 1965). According to this theory females'' wage rate is one of the most important determinants of fertility as the wage rate reflects the opportunity cost of having children. The prediction of a negative effect of female wages on fertility has been tested in a number of studies, but the results are diverging. Most of the studies using the hourly wage rate - as we do - are old ( Heckman and Walker 1990, Taşiran 1995, Merrigan and St.-Pierre 1998, Walker 2002). We contribute with new evidence based on registry data covering all Norwegian women born 1955–74 and a simultaneous hazard model of transitions to first, second and third birth. By dividing the women into four different groups based on year of birth (5 year birth cohorts), we can analyze whether the wage effect has changed over time among younger and older birth cohorts. So far we have found a U-shaped relationship between wages and the log hazard for all the four cohorts we are studying, however varying in strength and across parity. In transitions to first birth, most women are likely to be on the downward slope of the curve, implying that the wage effect is mainly negative. In transitions to second and third birth, most women are likely to be on the upward slope of the curve, where the wage effect is positive.

However, as women are located at different points on the U-shaped curve, it is hard to summarize the total effect of wages on womens'' fertility. The fact that the parameter estimates of the hazard model determines the effect on the log hazard and not the probability of have given birth at a particular age, complicates the interpretation of the parameter estimates further. To fully understand the relationship between wages and fertility, we also want to do simulations based on the estimated model. Then can calculate the relationship between the age of the woman and the probability of giving birth to the first child, the second child and the third child, respectively. Using an actual sample of women (we have register data) we can also calculate the number of women giving birth to one child, two children or three children.

The focus will be on the effect of wage changes on fertility. In the first simulation we want to study the effect of a general wage increase for all women. This type of simulation will provide us with knowledge about whether the effect of wage changes has changed across the four different cohorts we are analyzing. During the period covered by the sample there has been several expansions of family policies in Norway, and this might have changed the association between wages and fertility across cohorts. There has been an increase in paid parental leave from about three months in the mid 1970s to more than a year in 2009, the number of childcare centres has grown considerably, and a system of maximum prices has greatly reduced the prices in care centres.

In the second set of simulations we want the study whether the effect of a wage change on fertility vary across women in different parts of the wage distribution. Are women in the lower part of the wage distribution more sensitive to wage changes than women in the upper part of the distribution, and if they are, how large is the difference? To analyze that, we will divide the women into three groups according to wage, low wage women, medium wage women and high wage women. Then we can do simulations for each of the three groups where we increase the wage rate and study the effects for transitions to first, second and third birth, respectively. Also in these simulations we can compare the effects across different birth cohorts.

The main dataset covers all women living in Norway during the period 1974–2009. Here we observe the date of birth of all children, as well as the woman’s age and other relevant characteristics of the woman. A special feature of our approach is that we have access to data from the Norwegian Labour Force Survey for every year 1974–2009, i.e., for the same period as for the register data on fertility. Linking information about hours of work from this survey with information about actual wage incomes from the income registries, we can calculate wages as the fraction of wage incomes and actual hours of work, and estimate separate wage equations for each of the years 1974–2009. Based on the estimated equations we imputed year-specific hourly real wages (2005 price level) for all women in the study. By estimating the wage equations separately for each year, business cycle variations are allowed to influence wages and birth transitions.

Presented in Session 1235: Posters