Co-Parenting and Careers after Divorce

Katarina Kuske (Bocconi) — “Co-Parenting and Careers after Divorce,” presented at the NBER Summer Institute, Gender in the Economy, on 28 July 2026 (Day 1, afternoon). No discussant; questions from ten minutes in. Written from the 25 November 2025 job market paper.

There is a well-known literature on the co-location problem. Two people in a couple have two careers and one address. The address gets chosen, and on average it gets chosen to suit the higher earner, which makes the other one a tied mover or a tied stayer, and that costs her money. This is one of the more robust findings in the economics of the family.

Kuske’s paper is what happens when you dissolve the couple and the constraint doesn’t dissolve with it.

Joint physical custody — co-parenting, roughly fifty-fifty — is good for children; that is not much disputed. It also means the child has to be physically handed between two households on a schedule that repeats every week for a decade. Which means both parents have to stay near each other. Which means the co-location constraint that used to bind a marriage now binds a pair of ex-spouses, and it binds them in the same asymmetric way, for the same reason: the location was chosen during the marriage to suit his career, and after the divorce she is still standing in it.

As Kuske put it: this is not a gendered reform. The effect shows up as a gendered story because of household specialization.

The reform

The Netherlands passed a custody reform in late November 2008, effective 1 March 2009, establishing that a child whose parents held joint parental authority has the right to equal care and upbringing from both after divorce. A Supreme Court ruling in 2010 clarified that this made joint physical custody the prioriteitsmodel — the default preference, not an absolute rule.

Crucially, Dutch parents already had joint legal custody automatically before the reform. So this isolates physical custody, the actual living arrangement, from the legal-authority question that the closest American and Swedish papers are mostly about. Uptake among parents with young children rose 7.6 percentage points.

The design compares parents divorcing in the 18 months after the reform to those divorcing in the 18 months before, using Dutch administrative registers — monthly hours, earnings, employer sector and location — supplemented by the New Families in the Netherlands survey, which is where actual co-parenting status comes from, since the admin data cannot see who parents whom. Someone asked how “joint custody” is defined given that in the US it is very far from fifty-fifty in practice; the survey records which parent the child was with on each day of the month, and Kuske codes co-parenting as 40 to 60 percent.

The obvious threat — selection into divorce — was raised from the floor and has a decent answer: no jump in the number of divorces, no difference in observables pre and post, and an official Dutch Ministry of Justice review that checked whether the reform created new barriers to divorcing.

Mothers’ wages fall, and keep falling

Flat before, then a steady slide to about −1.2 log points by three semesters after. The headline intention-to-treat estimate is a 0.8 percent wage decline, which — scaled by the 7.6-point first stage — implies roughly a 10 percent wage loss for compliers. It is not a level shift; it is slower wage growth, which is why the line keeps drifting down rather than stepping down.

Fathers’ wages and hours are unaffected. Mothers’ hours fall temporarily, and the paper is scrupulous about why: it is largely that the control group’s hours rise, through overtime, rather than the treated group’s falling.

The mechanism is a map

If this is co-location, it should show up in geography, and it does. Treated ex-partners end up working about one kilometre closer to each other, against a mean post-divorce distance of 27.5 kilometres — and they are more likely to work in the same municipality, from a base where 52.9 percent already do. Mothers in the control group are the ones moving further away to better-paid jobs; treated mothers are not.

A kilometre sounds trivial until you notice it is an average across everyone, including all the people for whom the constraint never bound. What it is proxying for is the option — foreclosed — of taking the job two towns over.

The heterogeneity confirms the story rather than merely decorating it. The wage penalty concentrates among mothers who were secondary earners during the marriage and who were younger at divorce (below the sample median of 38). Those are exactly the mothers whose marital address was chosen for someone else’s career. Kuske noted that if you look at secondary earners who happen to be fathers, the effect is similar in magnitude — which is the cleanest possible demonstration that the mechanism is specialization, not sex.

What the compliers look like, and why it matters

The people who take up co-parenting are strongly positively selected: more educated, more likely to be employed, working more hours, with smaller gaps in hours between them, sharing childcare more equally before the divorce, and holding less conservative gender norms. Good moms and good dads, as Kuske put it — the people who can choose it and afford it. The compliers induced by the reform are slightly less positively selected than the always-takers, but still positively selected.

This is worth dwelling on because it inverts the usual worry. The estimate is not contaminated by dysfunctional families; it is drawn from the most functional ones. Which means the wage penalty documented here is what co-parenting costs when it is going well — among couples who chose it, who share childcare, who get along well enough to run a fifty-fifty schedule. The paper’s framing as an “efficiency cost of location constraints” is exactly right, and it is a cost being paid by the arrangement everyone agrees is best for the child.

The unmodelled channel

The best question came early and was not fully answered. Joint physical custody typically reduces the higher earner’s child support obligation, since the child is now living in both houses. And there is research suggesting fathers who are more involved shirk less on the support they do owe — the money reads as going to the child rather than to the ex-wife. So there are two income effects running in opposite directions, neither of which appears in a wage regression, and both of which change what the mother’s household actually has to spend.

Kuske does not have the child-support regime in her data. The paper’s outcome is the mother’s own hourly wage, which is the right object for a claim about careers and about the gender wage gap. But a reader wanting to know whether treated mothers are worse off overall cannot get that from here, and should not assume it.

(The tidy version: a policy designed to give children both parents works by fixing both parents to a point on a map. Divorce is supposed to be the thing that releases you from the compromise. It turns out that if you do the divorce properly — if you keep the child’s life intact, which is the thing we all agree you should do — the compromise outlives the marriage, and it is still costing the same person the same money.)