5 Estimated motherhood earnings losses for women with strong labour market attachment

View the most recent version.

In this section, we present our estimates of the earnings effects of childbirth and test the relevant hypotheses.²³ We focus on those mothers who gave birth once. In the next section, we discuss the robustness of our results with respect to those mothers who had one or two births and whether the effect of the first-born child is different from that of subsequent children.

5.1 Basic results

Table 4 contains the estimates for the 3,714 mothers who had one birth during the 1991-to-1996 period. The comparison group consists of the 20,992 women who did not give birth during the 1983-to-2004 period. The dependent variable in Panel 1 is the natural log of annual earnings, while that in Panel 2 is the annual earnings. We include the regression result in annual earnings in order to give the reader an idea of the dollar amount of the effects of childbirth.²⁴

First, our estimates show that the earnings differences between mothers and other women were not likely to be spurious because the differences were significant, even when the fixed effects were controlled for. Our results from both the fixed-effects model and the more general fixed-trend model suggest that the earnings differences were close to 40%—over $11,000—in the year of childbirth and 30%—close to $8,000—in the first post-childbirth year. These were mostly the results of new mothers' temporary withdrawal from the labour market and would be partially compensated by the Employment/Unemployment Insurance benefits. We investigate the effects of these benefits on the earnings trajectories of mothers in the next subsection.

However, childbirth did not only affect the earnings of the mothers during the year of childbirth or the first post-childbirth year: on the contrary, the earnings difference continues to be there over a number of post-childbirth years. For example, the fixed-effects model estimates indicate that, from the second to the ninth post-childbirth years, the annual earnings differences between mothers and the comparison group ranged from 5%—around $1,300—to 10%—around $2,700—with lower earnings penalties occurring in the years further after the child is born. The fixed-trend model estimates point to a similar conclusion, albeit the estimated magnitude of the motherhood earnings penalty was smaller than those under the fixed-effects model.

Table 4
Basic estimates, fixed-effects and fixed-trend models

But the effects of childbirth on the earnings of the mothers did not seem to last forever. Under the fixed-effects specification, motherhood earnings losses declined from 10% to 5% from the second to the ninth post-childbirth years; while the estimates from the more general fixed-trend model indicate that the earnings gap between mothers and non-mothers amounted to 8% in the second post-childbirth year, and they declined to 5% in the sixth year, and 3% in the seventh year. The estimate for the seventh year was only significant at the 10% level. After the seventh year, the estimates were no longer significant. These findings contradict the notions that human capital losses due to childbirth permanently reduce the earnings of a mother, and that mothers who interrupted their careers as paid employees for childbearing and rearing can never entirely regain their potential earnings.

The estimation results also suggest that the annual earnings of Canadian mothers did not decline during the years before they gave birth. On the contrary, they continued to increase, at least in the year immediately before childbirth. The fixed-effects estimates show that the earnings in the pre- childbirth year increased by 3%, while the fixed-trend model indicates a close to 4% increase. These cast doubt on the prediction that women whose earnings declined—for example, due to bad job

matches—may choose to become mothers.²⁵ Hence, our results do not support the endogenous motherhood hypothesis.

In addition, as noted before, our estimated earnings differences between mothers and other women were lower under the fixed-trend model than under the fixed-effects model. We also notice that the further a post-childbirth year is from the year of childbirth, the more the earnings differences between mothers and other women can be explained by the inclusion of the individual-specific earnings growth trend. For example, in the second post-childbirth year, the earnings effect of 10% (from the fixed-effects model) was reduced to 8%, once the individual-specific earnings growth trend was taken into consideration (fixed-trend model). But in the seventh post-childbirth year, the corresponding earnings effects were reduced from 7% in the fixed-effects model to 3% in the fixed- trend model.

Figure 4
'Actual' and estimated earnings profiles

The above result is illustrated in Figure 4, where fixed-trend estimates of the earning effects of childbirth are lower than those of the fixed-effects estimates; and, the further a post-childbirth year is after the year of childbirth ( t = 0), the fixed-trend and the fixed-effects estimates are further apart. Hence, our results suggest that women who become mothers may have flatter intrinsic age–earnings profiles than other women.

Given that the estimated effects of childbirth on the earnings of the mothers declined over the post- childbirth years, one would be tempted to confirm the work-effort hypothesis and to reject the human-capital hypothesis. Under the former, the earnings losses of the mothers would decline over the post-childbirth years while, under the latter, the negative effects would stay constant. Unfortunately, our data do not allow us to directly test these hypotheses, because some key variables are not available: for example, we do not have data on work effort and hours worked. If a mother reduces her working hours upon giving birth, and increases them gradually as the child grows older while keeping her work effort constant, we would also be able to obtain the current estimates.

In summary, the basic results indicate that there exist some substantial earnings gaps between Canadian mothers—at least for those who had strong labour market attachment—and their non- mother counterparts, and the effects were unlikely spurious. But the gaps declined over the post- childbirth years and, according to the flexible fixed trend model, the differences would disappear seven years after childbirth.

5.2 Effect of Employment Insurance benefits

As mentioned above, employed Canadian mothers who temporarily withdrew from the labour market due to childbirth were eligible for compensations in the form of Employment Insurance (EI) benefits. From 1991 to 1996, mothers were eligible to receive the benefits for a maximum of 25 weeks, and the statutory replacement rates ranged from 55% to 60%.²⁶ An immediate question is how EI benefits affect our estimated effects of childbirth on earnings.

Between mothers and non-mothers, since only the former could receive the maternity benefits during a certain period of time surrounding childbirth, it is straightforward to answer the above question—for example, by subtracting the maternity benefits received from the estimated earnings losses. But in the Longitudinal Worker File, we do not know when a mother started to receive the benefits and if all of the benefits she received were related to maternity leave. This would be the case when a mother might experience other job separations not related to the childbirth; in our narrow sample, we only assume mothers did not experience any permanent layoffs.

Given the limitations of the data, a simple way to address the above question is to add the EI benefits received to the total employment earnings for all women—mothers and women from the comparison group—in all years of observation. In other words, we shall define a new 'earnings' concept that equals the sum of employment earnings and EI benefits, and then re-estimate the basic model. Results from the fixed-trend models are contained in Table 5.²⁷

Table 5
Effect of Employment Insurance benefits, fixed-trend model

The results show that the earnings gaps between mothers and other women dropped substantially in the year of childbirth and the year thereafter, once the maternity benefits are taken into consideration. In particular, our basic estimates (Table 4) suggest that the total earnings effects in the year of childbirth and the first year after were close to $19,000. But once the benefits were taken into account, as Table 5 indicates, the corresponding estimates decreased to about $10,000.

The estimated-earnings effects from the second to the eighth post-childbirth years were also 1 to 2 percentage points lower under the new earnings definition than previously estimated. In terms of dollar amounts, the total earnings gap was close to $9,000 over the seven post-childbirth years. In contrast, when EI benefits were not included, the total earnings gap over these seven years amounted to about $10,000 (Table 4). Hence, the EI benefits further reduced earnings effects of childbirth by about $1,000 for these post-childbirth years. Taken together, the total motherhood earnings gap—over all the post-childbirth years—was reduced from $29,000 to $19,000, due to the maternity benefits.²⁸

5.3 Do mothers who change employers lose more?

We next estimate the model controlling for the fact that some mothers worked for different employers after giving birth. Since those who worked for the same employers after giving birth would lose little of their firm-specific human capital or their job tenure, we would expect their earnings drops to be significantly lower than those who changed employers after having given birth.²⁹

Table 6
Effect of employer changes

This was indeed the case as shown by the estimates in Table 6. Estimates from the fixed-effects model show that the negative effects of childbirth on the earnings of mothers who worked for the same employer amounted to from 3% to 6% during the post-childbirth years, much less than the losses incurred by mothers who changed employers. But once we allow for individual specific- earnings growth paths, the effects were only significant in the year of childbirth and in the first two years thereafter, and the estimate for the second post-childbirth year was only marginally significant.

While the finding that mothers who stayed with the same firm incurred little earnings drops and those who changed employers incurred significantly higher earnings drops is consistent with the predication of firm-specific human-capital hypothesis, the result can also be interpreted as one that is due to the losses of earnings premium that had resulted from good job matches. Alternatively, it is possible that when a mother changes employers, she may change from a full- time job to a part-time one, or she may change from a well-paid job to aless well-paid one in order to facilitate child rearing and caring.

 

²³ The model is estimated in two steps. In the first step, we regress annual earnings on a series of year dummies and the provincial specific unemployment rates of female workers 15 years old and over (the data were retrieved from CANSIM, Table 282-0086). This purges the effects of the overall earnings growth trend and the provincial specific business cycle on earnings. In the second step, we regress the residuals from the first step on the quadratic term of age and year of childbirth indicators.

²⁴ Regression in level for other specifications will not be included hereafter but they are available from the author.

²⁵ Ejrnaes and Kunze (2004) discussed this notion of endogenous motherhood.

²⁶ The number of weeks of maternity benefits was extended to 50 weeks as of January 1, 2001.

²⁷ Fixed-effects models were also estimated (available upon request). The estimated earnings losses were slightly higher than for those under the fixed-trend models. For example, in the year of childbirth and the year after, the fixed-effects estimates were around -0.18 and significant.

²⁸ On top of the maternity benefits provided by the federal government, some employers offer their workers supplementary benefits for maternity and parental leaves. Although we do not have any information on the percentage of the supplementary benefits, about one third of employees (regardless of gender) reported that their workplaces offer such benefits (Workplace and Employee Survey 1999, 2001 and 2003). Hence, one third of Canadian mothers would have an earnings gap less than $10,000, depending on how much their employers topped up the maternity benefits.

²⁹ An important reason for a new mother to change employer is to facilitate child caring and the schooling of the child. This generally occurs when the child is aged 5 or younger. To reflect this, we define a dummy variable that equals 1 if a mother worked for the pre-childbirth employer in at least five consecutive years after giving birth. We interact this dummy variable with each of the years of childbirth indicators ( D_it^k ).