Print Post
  • In a new working paper, Peter Q. Blair and Benjamin Posmanick conclude that parental leave can explain a whopping 94% of the slowdown in gender wage convergence. Tweet This
  • Men and women are unique, and all sorts of societal forces affect them differently; assuming a linear march to complete equality strikes me as a stretch. Tweet This
  • Like the authors, I find a big drop in the gender wage gap in the 1980s and early 1990s, followed by a slower decline thereafter. Tweet This
Category: Work-Family

Is parental leave backfiring? One reason for the policy is to keep women attached to their jobs after giving birth, so they can take some time off but then return a bit later. Under federal law, all parents of both genders are entitled to 12 weeks of unpaid leave, and some states have been giving paid leave as well.

Things could actually work in the opposite way, however. Women are more likely than men to take leave when it’s available. Employers might decline to hire them to avoid having to deal with long absences, the time off might set women behind at work or establish expectations at home that they’ll take the dominant role in child care, and some women, heaven forbid, might decide they enjoy being full-time moms.

I wrote about one study with findings in this vein back in 2019; it dealt with paid leave in California and found the policy reduced first-time new mothers’ “employment by 7 percent and lowered annual wages by 8 percent six to ten years after giving birth.” And now there’s a working paper from Peter Q. Blair and Benjamin Posmanick, released through the National Bureau of Economic Research, with even more striking results.

Blair and Posmanick point out that the gender wage gap shrank quickly from the late 1970s through the early 1990s, starting at roughly 40% but then falling by about a percentage point a year. After 1993, it fell at a far slower rate, about a third of a point per year. If the gap had continued shrinking at its faster earlier rate, it would have disappeared entirely circa 2017.

As it happens, 1993 was the year of the Family and Medical Leave Act, the federal law ensuring most workers access to unpaid parental leave. Coincidence? The authors think not.

They test the idea with a series of intense statistical analyses of the Current Population Survey’s Annual Social and Economic Supplement, often focusing on 12 states, plus D.C., that enacted parental leave before the federal law, with Massachusetts starting the trend in 1972. (Some of these states only offered maternity leave, not paternity leave.) They conclude that parental leave can explain a whopping 94% of the slowdown in gender wage convergence. They also find it works in large part by increasing men’s wages rather than by reducing women’s.

This is an interesting paper with a provocative thesis, but there are a few reasons to be skeptical the effect is really this big. One involves the implication that the gender wage gap could have ended in 2017. This requires a pretty strong assumption that just because the wage gap shrank by a point a year in the 1980s, one would naturally expect that to continue until the gap hit zero (at which point, presumably, women’s one-point-per-year advance on men would simply stop?). Men and women are unique, and all sorts of societal forces affect them differently; assuming a linear march to complete equality strikes me as a stretch. It seems far more likely to me that convergence was bound to slow down at some point.

My second criticism is a bit more technical: Their models measure the gender gap over time for different demographic groups and states, but they do not control for changes to the gender gap that occur in all states regardless of whether they have parental leave. In more technical terms, the models predicting wages don’t include a sex-by-calendar-year interaction, and therefore might attribute changes to the gender gap that occurred everywhere, regardless of parental-leave laws, to the laws themselves.

I contacted the authors to ask about this, and Posmanick responded: 

We do not include these interactions due to concerns about asking too much of the data because we have about 13 events at the state level, and several occur in the same calendar year. Therefore, we are concerned the additional interactions will absorb too much of the meaningful variation.

And third, I did a simple test of the theory myself. I downloaded the numbers from the same source (IPUMS) and processed the data essentially as they describe in the paper.1 (My wage-gap estimates are generally slightly lower than theirs but follow the same trends.) Here is what the overall gender gap looks like, dividing the median female wage by the median male wage and subtracting from one. Like the authors, I find a big drop in the 1980s and early 1990s, followed by a slower decline thereafter.

Then, I split the country into two groups.

One group of states implemented maternity leave between 1972 and 1988: Massachusetts, Connecticut, Washington, California, Minnesota, Rhode Island, Maine, Oregon, Tennessee, and Wisconsin. This group was basically under constant assault from the policy in the early part of the chart, as more and more states implemented the law, some of which gave the benefit only to women. The entire group had the law five years before the federal policy set in. So, if maternity leave fundamentally changes the trend in the gender wage gap, this group should see stagnation set in early.

The second group comprises all the states that never implemented leave by themselves but received it in 1993 with the federal law. This group should have especially strong wage-gap convergence until the early 1990s and then an especially abrupt shift thereafter. (A few states that implemented parental leave in the early 1990s just before the federal law—New Jersey, Vermont, and D.C.—are simply dropped to keep some separation between the two groups in terms of when they got leave.)

Here are the gaps in median wages over time for the 1972-1988 and 1993 groups:

That doesn’t look too dramatic. Maybe one could argue that the earlier-treated group flagged a little bit in the mid-1980s, or that the 1993 cohort stalled slightly worse in the mid-1990s. And these simple trend lines don’t account for other confounding variables.

But in general, both groups have the same trend: A big drop through the early 1990s, a smaller drop thereafter. If I run simple regressions to get the pre-1993 vs. 1993-and-after trend lines, both groups saw their gender gap drop by roughly a point per year beforehand and 0.3-0.4 points after.

I can buy that parental leave reduces women’s economic participation and wages to some extent, and I hope this paper receives a lot of careful attention. For the time being, though, I’m skeptical that family leave is practically the sole reason for the slowdown in gender wage convergence.

Robert VerBruggen is an IFS research fellow and a fellow at the Manhattan Institute.

1. Like the authors, I restrict the sample to whites and blacks with 35 or more hours per week, positive wage income (and more than zero weeks worked), and ages 18-65 in the year the income was earned (ages 19-66 in the survey year); and calculate wages by dividing wage income by the product of weeks worked and typical weekly hours. I start my analysis one year after theirs; in the 1976 survey year (pertaining to 1975 wages), the CPS data identify many states in groups rather than individually. Hourly wages are calculated as wage income divided by the product of usual hours and weeks worked. Since I am working only with medians and the ratio of male to female wages on a year-by-year basis, however, there’s no need to trim the top and bottom 1% of observations or adjust for inflation.