Heida M. Sigurdardottir Dept. of Neurosci. Brown Univ. Box G-LN Providence, RI 02912

A way of promoting gender equality? Sharing of parental leave between fathers and mothers

Olof Gardarsdottir and Heida Maria Sigurdardottir

Some more information (in addition to information on the poster):

First we performed linear regression with various background variables, such as education, that might covary with the father's income group. Each of these variables explained a portion of the variability in the leave taken. There was still a lot of variance left unexplained. For each data point (one child) we took the residuals, i.e. the difference between the days of the father's leave predicted by the model and the days of leave each father actually took (the unexplained variability), and used them in a follow-up analysis. The residuals are still in days of leave, but now they are days of leave above those predicted by the linear regression model, i.e. days that we have yet not explained. We then see if we can account for some of that variance in our follow-up analysis. The residuals should, for the most part, be free of the effects of the background variables, so our measure of the effect of income should be more "pure".

For each month and for each income group, we calculated the mean residual. In this secondary analysis, each data point is now not a single child, but the average child in each month, i.e. how many days, above those predicted, does a typical dad in that month and that income group take? As can be seen from the scatter plot, the mean residuals (days above predicted) seem to vary with both time and income group. To quantify this, we fit four second order polynomials, one for each income group. We chose polynomials instead of a linear fit because we expected an event (economic recession/laws) to change the nature of the relationship between income and days above predicted (i.e. we expected a curvilinear relationship).

Fits indicate that all four income groups, but especially the highest income group, have gone from first taking more and more leave over time but then starting to take less and less leave in more recent years. The question is, would this have happened anyway even if there had been no economic crisis and no change of the law? To get a sense of this, we fit polynomials for a) all data (longest period), b) data up until the change of the law (maximum payment during parental leave was lowered to a sum below the mean wages), c) data up until the economic crash (shortest period).

If we look at the data from before the crash (c), there is absolutely no indication that the days of leave taken would start to diminish at some time point. All income groups would have been expected to go up, if anything. The question that still remained was whether there was something non-specific about the economic recession that was driving the change or whether it was actually the change in the law that was the main factor of change. We therefore compared (a) with (b) to see if the change had already started to happen before the changing of the law. This was the case for the lowest income group, but not the other three higher income groups. The lowest income group is also the group not really affected by the change in the law. Therefore, non-specific effects of the economic crisis might be driving the change in this group. The other groups do not really change from growth to decline unless data from AFTER the law change are included. The drop in these groups might therefore be mainly driven by the change in the law.