You are here

Marriage, Divorce & Child Outcomes: The Role of Public Policies

December 11, 2019
How do family laws that govern divorce, custody and child support mold the future of children, in their youth and adulthood? Joseph Mullins, U of M assistant professor of economics, has been studying that very question in cooperation with Christopher Flinn, New York University professor of economics and Meta Brown, Ohio State University associate professor of economics. 

You’re a native of Australia, a long way from your original home. What led you to the faculty at the University of Minnesota?

I first came to the states to do my PhD at NYU. That was a little over 10 years ago. After that I spent two years at the University of Western Ontario, in Canada. If you track my path, I’ve been consistently climbing further north. I’m looking for any opportunity in Saskatchewan or maybe Alaska or maybe the North Pole, ultimately.

Coming here gave me the unique opportunity to interact with some of the world’s best economists. When it comes to thinking about economic models, there aren’t many better places to be than the University of Minnesota.

You’re looking at how public policies influence how members of a family fare – especially children. Why is that of interest to an economist?

There’s a body of evidence that early childhood environments are crucial in shaping skills of adults. We might divide these skills into two categories – cognitive and non-cognitive. But both of these features turn out to be important in shaping your value in the labor market, your propensity to commit crime, your health.

How did you go about your study?

The specific topic I’m studying is about laws governing marital dissolution. These laws shape, fundamentally, decisions about who gets divorced and when.

Those decisions are incredibly influential in two dimensions. The impact it has on household resources and time use decisions. How much time the parents decide to spend with their kid and in the labor market. We know this a channel through which policy can change the skill formation of both young children and adolescents.

Joseph Mullins
Professor Joseph Mullins

What’s the source and size of the panels you examined?

We’re making use of a publicly available data set known as the Panel Study of Income Dynamics started in 1968 as a representative sample of U.S. households. It tracks the children of the original household members.

You’ve got 51 years of data on outcomes?


In 1997, they introduced something very valuable for researchers like me. They introduced something called the Child Development Supplement, which collects developmental indicators for children in PSID households. Psychologists have designed these tests to ensure that they’re measuring something really robust about skills. It will give you a pretty darned good sense of who’s doing well in school. Whatever their aptitude scores at a young age usually are predictive of their later labor market skills and their health, and so forth.

How can you distinguish between cause and effect? Roosters crow before sunrise but that doesn’t make the sun rise.

Staying married or divorced is the result of a lifetime of decision making.

Our solution is to model that selection. To use the panel data that suggest how much time parents choose to spend on their children, how much they invest. To tease out the correlations from the outcomes. To determine whether it’s the cock crowing or the sun rising.

How do you tease it out?

You might imagine it’s a fiendishly difficult problem – and it is.

For example, we’ll look at exposure to divorce. Instead of just comparing children of divorce to children of intact households, we’ll look at things that might be more informative comparison. To look at children whose parents have been divorced for five years versus children whose parents have been divorced for one. We'll also try to use differences across states and over time in prevailing marital law. This would be like if the farmer removed the proverbial rooster randomly before each sunrise. Then we're really in business when it comes to these causal questions.

In some states, divorce is a contest of warring parties. In other states, a divorce can proceed with a spouse simply telling their husband or wife “We’re done.” 

At the beginning of 20th Century, most states had a “mutual consent” divorce law in which both parties would have to consent to a divorce. In the 1960s and 1970s, we saw a transition in a lot of states to “unilateral” or “no-fault” divorce. I no longer had to prove that my spouse was at fault in order to get out of the marriage.

Do you have any sense on why those changes would make a difference?

As soon as we decide that a household consists of two independent decision-making agents, as opposed to two individuals who are fused to become one, the circumstances have changed.

We also see that in the years prior to a divorce, both
husbands and wives start to spend less time with their
child, and wives start to work more.

Husband and wife may not agree on the model of a new car or what color to paint the living room.


We can’t say how a couple is going to decide to split resources. Maybe we can assume that they’re not going to leave something on the table, but one implication of this efficient household is that the prevailing divorce law should have no effect on their decision to get divorced or not. This is not what the data suggests. We also see that in the years prior to a divorce, both husbands and wives start to spend less time with their child, and wives start to work more. We have found it hard to write down a model in which this is the result of efficient decision-making.

A family that stays together isn’t necessarily better for the kid. In other words, this isn’t an indictment of divorce. 

That’s correct. Crucially, we find that a bad marriage could be far worse than a dissolved marriage for developmental outcomes for the children.

Again, the laws may determine that. If we live in a world where divorce requires mutual consent, it could be the case that what we’re really doing is just keeping bad marriages intact. These are very complicated questions. Our aim is to use a model to untangle them all. 

How big is your sample?

We have a representative sample of approximately 1,500 children to examine.

Does it matter if a child is three or nine or 15 when a marriages dissolve?

It sure does. A crucial element is “How important is time investment” at different stages of childhood? The rule of thumb, from previous studies, is that time invested is much more productive for young children, less than age five. The closer and closer you get to birth, the more and more important this time is.

How does the amount of parenting time – and intensity of attention – differ between married and divorced parents?

We know that divorced households have fewer resources to draw upon in order to invest in kids. But if it’s really true that it’s worse for children to have their parents divorce, we should probably see that in the data.

What we find, in fact, is a relatively flat profile with respect to exposure to divorce. The problem is there’s no such thing as a perfect comparison between married and divorced. There are all these other factors. There are good marriages and bad marriages.

If outside (legal) incentives are shifting to keep good marriages intact, that’s good for kids. If they’re shifting to keep bad marriages intact, that’s bad for kids.

How strong a link can you find between child test scores and the earnings and marital status of the parents?

Very strong. The question is how much of that is causal as opposed to other factors – parental ability and so forth.

I assume that as income rises, so do the test scores?

Yes. There’s literature, that runs parallel to ours, trying to understand the roles that income can play in shaping child outcomes. 

How does government figure in the questions you’re asking?

Think about the 1996 welfare reform. It tried to create a lot of incentives to foster household stability – to keep the man in the household, to keep couples together and stay married.

Does it matter which parent prevails, in terms of a son or daughter’s long-term educational performance?

If the government’s goal is just to decide on an arrangement that’s best for kids, then it depends on the stage in the lifecycle that the child is in. There’s modest evidence to suggest that early in life the mother’s time is much more important and that, by adolescence, the father’s time becomes more important.

You might think that it’s best to come up with a custody policy that’s more dynamic. Think of a legal framework that favors mothers with kids of a certain age and favors equal custody (between mothers and fathers) for kids of an older age.

The Heller-Hurwicz Q&A series shares exciting preliminary research findings from University of Minnesota economists. The text has been edited for clarity and length.